{ "version": "1.1.5", "last_updated": "2026-10-08", "description": "", "matches": [ { "match_id": "arena-nvidia-500b-decentralized-compute", "date": "2026-08-11", "type": "content", "topic": "Nvidia's $500B AI Compute Push Is the Bull Case for Decentralized Networks", "blog": "niteagent", "slug": "nvidia-500b-ai-compute-decentralized-networks", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.1, "word_count": 1275 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 3.5, "word_count": 1124 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "failed", "score": 0.0, "word_count": 0, "error": "API error: openrouter/poolside/laguna-s-2.1 is not a valid model ID" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "score_a": 8.1, "score_b": 7.8, "winner": "A" }, { "model": "glm-5.2", "provider": "zai", "family": "glm", "qa_score": 8.4, "qa_pass": true } ] }, { "match_id": "arena-handoffs-vs-supervisor", "date": "2026-08-10T16:04:00+00:00", "type": "deep-dive", "topic": "Handoffs vs. Supervisor: Choosing Your Multi-Agent Control Flow", "blog": "niteagent", "slug": "handoffs-vs-supervisor-multi-agent-control-flow", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 7.8, "word_count": 1506 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 6.8, "word_count": 1163 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "failed", "score": 0.0, "word_count": 0, "error": "API error: openrouter/poolside/laguna-s-2.1 is not a valid model ID" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "score_a": 7.8, "score_b": 6.8, "winner": "A" }, { "model": "glm-5.2", "provider": "zai", "family": "glm", "qa_score": 8.6, "qa_pass": true } ] }, { "match_id": "arena-bounded-autonomy", "date": "2026-08-10T11:25:00+00:00", "type": "content", "topic": "Bounded autonomy: a field guide to the wallet-level guardrails now shipping for AI agents", "blog": "niteagent", "slug": "bounded-autonomy-guardrails-ai-agents", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.6, "word_count": 1288 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 4.5, "word_count": 1119 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "failed", "score": 0.0, "word_count": 0, "error": "API error: openrouter/poolside/laguna-s-2.1 is not a valid model ID" } ], "judges": [ { "model": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score_a": 8.6, "score_b": 6.8, "winner": "A" } ] }, { "match_id": "daily-20260808-0002", "date": "2026-08-08T16:00:00+00:00", "type": "deep-dive", "topic": "MCP's Broken Trust Model: Tool Poisoning, Rug Pulls, and the New Threat Landscape", "blog": "niteagent", "slug": "mcp-broken-trust-model-tool-poisoning-attacks", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.8, "word_count": 1874 }, { "model": "mimo-v2.5", "rank": 3, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.4, "word_count": 1960 }, { "model": "poolside-laguna-s-2.1", "rank": 2, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.5, "word_count": 1696 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.8, "score_b": 8.4, "score_c": 8.5, "winner": "A" } ] }, { "match_id": "daily-20260808-0001", "date": "2026-08-08T06:00:00+00:00", "type": "content", "topic": "AI Agents Got Bank Accounts This Week", "blog": "niteagent", "slug": "agent-payments-vs-agent-tokens-aug2026", "winner": "mimo-v2.5", "results": [ { "model": "deepseek-v4.1-flash", "rank": 3, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.0, "word_count": 1750 }, { "model": "mimo-v2.5", "rank": 1, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.7, "word_count": 1608 }, { "model": "poolside-laguna-s-2.1", "rank": 2, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.2, "word_count": 1616 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.0, "score_b": 8.7, "score_c": 8.2, "winner": "B" } ] }, { "match_id": "daily-20260807-0002", "date": "2026-08-07T16:00:00+00:00", "type": "deep-dive", "topic": "OpenTelemetry GenAI Semantic Conventions: The Agent Tracing Standard", "blog": "niteagent", "slug": "otel-genai-agent-trace-field-guide", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 9.4, "word_count": 2896 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.6, "word_count": 2679 }, { "model": "deepseek-v4.1-flash", "rank": 3, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.4, "word_count": 2631 } ], "judges": [ { "model": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score_a": 8.4, "score_b": 9.4, "score_c": 8.6, "winner": "B" } ] }, { "match_id": "daily-20260807-0001", "date": "2026-08-07T06:00:00+00:00", "type": "content", "topic": "Cloudflare Wallets and x402: How AI Agents Are Getting Their Own Payment Rails", "blog": "niteagent", "slug": "cloudflare-wallets-x402-ai-agents-payment-rails", "winner": "mimo-v2.5", "results": [ { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.5, "word_count": 1442 }, { "model": "mimo-v2.5", "rank": 1, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 9.0, "word_count": 1169 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.8, "word_count": 1456 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.5, "score_b": 9.0, "score_c": 8.8, "winner": "B" } ] }, { "match_id": "daily-20260805-0001", "date": "2026-08-05T06:00:00+00:00", "type": "content", "topic": "Why Agent Tokens Keep Dying — And What Bittensor Got Right", "blog": "niteagent", "slug": "why-agent-tokens-keep-dying-elizaos-bittensor", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.17, "word_count": 1705 }, { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 7.67, "word_count": 1668 }, { "model": "deepseek-v4.1-flash", "rank": 3, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 7.67, "word_count": 1825 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 7.67, "score_b": 8.17, "score_c": 7.67, "winner": "B" } ] }, { "match_id": "daily-20260804-0001", "date": "2026-08-04T06:00:00+00:00", "type": "content", "topic": "The EU AI Act Just Went Live — What It Means for Your AI Trading Bot and Agent Deployment", "blog": "niteagent", "slug": "eu-ai-act-ai-trading-bots-guide", "winner": "poolside-laguna-s-2.1", "results": [ { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 9.0, "word_count": 1990 }, { "model": "mimo-v2.5", "rank": 3, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.6, "word_count": 1450 }, { "model": "poolside-laguna-s-2.1", "rank": 1, "provider": "openrouter", "family": "poolside", "status": "success", "score": 9.2, "word_count": 1966 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 9.0, "score_b": 8.6, "score_c": 9.2, "winner": "C" } ] }, { "match_id": "daily-20260803-0002", "date": "2026-08-03T16:00:00+00:00", "type": "content", "topic": "MCP Just Went Stateless: Inside the 2026-07-28 Spec Rewrite", "blog": "niteagent", "slug": "mcp-stateless-spec-rewrite-2026-07-28", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.8, "word_count": 2444 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.2, "word_count": 2398 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.6, "word_count": 2594 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.8, "score_b": 8.2, "score_c": 8.6, "winner": "A" } ] }, { "match_id": "daily-20260803-0001", "date": "2026-08-03T06:00:00+00:00", "type": "content", "topic": "AI Agents Moved $100M in Two Weeks — Inside the Agentic Finance Stack", "blog": "niteagent", "slug": "agentic-finance-stack-2026", "winner": "poolside-laguna-s-2.1", "results": [ { "model": "deepseek-v4.1-flash", "rank": 3, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 7.0, "word_count": 1340 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 7.0, "word_count": 1400 }, { "model": "poolside-laguna-s-2.1", "rank": 1, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.0, "word_count": 2119 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 7.0, "score_b": 7.0, "score_c": 8.0, "winner": "C" } ] }, { "match_id": "daily-20260802-0001", "date": "2026-08-02T06:00:00+00:00", "type": "content", "topic": "PayBox and x402: The Dawn of Agentic Crypto Payments", "blog": "niteagent", "slug": "moonpay-paybox-agentic-payments-x402", "winner": "poolside-laguna-s-2.1", "results": [ { "model": "deepseek-v4.1-flash", "rank": 3, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.1, "word_count": 1096 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.7, "word_count": 1204 }, { "model": "poolside-laguna-s-2.1", "rank": 1, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.7, "word_count": 1196 } ], "judges": [ { "model": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score_a": 8.1, "score_b": 8.7, "score_c": 8.7, "winner": "C" } ] }, { "match_id": "daily-20260801-0001", "date": "2026-08-01T06:00:00+00:00", "type": "content", "topic": "The Agentic Money Stack Is Shipping — And Its Failure Modes Are Now Public", "blog": "niteagent", "slug": "agentic-money-stack-failure-modes", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.5, "word_count": 2177 }, { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.0, "word_count": 2000 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.0, "score_b": 8.5, "winner": "B" } ] }, { "match_id": "daily-20260731-0001", "date": "2026-07-31T06:00:00+00:00", "type": "content", "topic": "NEAR Protocol stake-to-pay AI inference: locked tokens as metered compute credits", "blog": "niteagent", "slug": "near-stake-to-pay-ai-inference", "winner": "mimo-v2.5", "results": [ { "model": "mimo-v2.5", "rank": 1, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.8, "word_count": 1156 }, { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.6, "word_count": 1097 } ], "judges": [ { "model": "mimo-v2.5", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.6, "score_b": 8.8, "winner": "B" } ] }, { "match_id": "daily-20260729-0001", "date": "2026-07-29T10:28:00+00:00", "type": "content", "topic": "AI-crypto stack maturation: regulatory clarity, token infrastructure, autonomous agent economy", "blog": "niteagent", "slug": "ai-crypto-stack-matures-regulatory-clarity-token-infrastructure-2026", "winner": "mimo-v2.5-pro", "results": [ { "model": "mimo-v2.5-pro", "rank": 1, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.3, "word_count": 2009 }, { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 7.8, "word_count": 2082 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 7.8, "score_b": 8.3, "winner": "B" } ] }, { "match_id": "daily-20260728-0001", "date": "2026-07-28T06:00:00+00:00", "type": "content", "topic": "AI agent payment infrastructure stack: x402, Visa stablecoins, AWS Bedrock, BNB Chain L1", "blog": "niteagent", "slug": "ai-agent-payment-infrastructure-stack-architecture-2026", "winner": "mimo-v2.5-pro", "results": [ { "model": "mimo-v2.5-pro", "rank": 1, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.1, "word_count": 1650 }, { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 7.2, "word_count": 1640 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 7.2, "score_b": 8.1, "winner": "B" } ] }, { "match_id": "daily-20260727-0001", "date": "2026-07-27T17:57:32+00:00", "type": "content", "topic": "AI agents crossed 100M crypto payments on Base — the agentic finance stack is live", "blog": "niteagent", "slug": "ai-agent-payments-100-million-x402-base-2026", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.1, "word_count": 1009 }, { "model": "mimo-v2.5-pro", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 7.2, "word_count": 1007 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.1, "score_b": 7.2, "winner": "A" } ] }, { "match_id": "daily-20260726-0009", "date": "2026-07-26T18:13:37.630003+00:00", "type": "content", "topic": "6 System Design Mistakes That Kill AI Agents in Production", "blog": "niteagent", "slug": "6-system-design-mistakes-that-kill-ai-agents-in-production", "winner": "mimo-v2.5", "results": [ { "model": "mimo-v2.5", "rank": 1, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.55, "word_count": 1251 }, { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.25, "word_count": 1251 } ], "judges": [ { "model": "codestral-2508", "provider": "mistral", "family": "codestral", "score_a": 7.7, "score_b": 8.2, "winner": "B" }, { "model": "nemotron-3-ultra-550b-a55b", "provider": "openrouter", "family": "nemotron", "score_a": 8.8, "score_b": 8.9, "winner": "B" }, { "model": "llama-3.3-70b-versatile", "provider": "groq", "family": "llama", "score_a": null, "score_b": null, "winner": "N/A", "status": "failed", "error": "model_not_found on cerebras; rerun blocked by compression-model config error" }, { "model": "qwen3.6-27b", "provider": "groq", "family": "qwen", "score_a": null, "score_b": null, "winner": "N/A", "status": "failed", "error": "subagent blocked by compression-model context-length config error" } ] }, { "match_id": "daily-20260723-0001", "date": "2026-07-23T17:39:18.152947+00:00", "type": "content", "topic": "Multi-Agent Failure Modes: 7 Patterns That Break Production Systems", "blog": "niteagent", "slug": "multi-agent-failure-modes-7-patterns-that-break-production-systems", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.83, "word_count": 1938 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.1, "word_count": 1743 } ], "judges": [ { "model": "llama-3.3-70b-versatile", "provider": "cerebras", "family": "llama", "score_a": 8.9, "score_b": 7.5, "winner": "A" }, { "model": "qwen3-32b", "provider": "groq", "family": "qwen", "score_a": 8.5, "score_b": 7.7, "winner": "A" }, { "model": "codestral-2508", "provider": "mistral", "family": "codestral", "score_a": 8.9, "score_b": 8.8, "winner": "A" }, { "model": "nemotron-3-ultra-550b", "provider": "openrouter", "family": "nvidia", "score_a": 9.0, "score_b": 8.4, "winner": "A" } ] }, { "match_id": "weekly-20260721-0001", "date": "2026-07-21T10:08:07.567145", "type": "content", "topic": "How to Set Up CrewAI for Multi-Agent Workflows (2026)", "blog": "niteagent", "slug": "how-to-set-up-crewai-for-multi-agent-workflows-2026", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.67, "word_count": 1204 }, { "model": "mimo-v2.5-pro", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.0, "word_count": 1609 }, { "model": "deepseek-v4", "rank": 3, "provider": "nexum", "family": "deepseek", "status": "failed", "error": "DialagramRouter upstream unavailable" } ], "judges": [ { "model": "llama-3.3-70b-versatile", "provider": "groq", "family": "llama", "role": "judge" }, { "model": "gpt-oss-120b", "provider": "cerebras", "family": "gpt", "role": "judge" }, { "model": "mistral-small-2506", "provider": "mistral", "family": "mistral", "role": "judge" } ], "winner_content_path": "/tmp/arena-outputs/weekly-20260721-0001/deepseek-deepseek-v4-flash.mdx" }, { "match_id": "guide-20260721-0300", "date": "2026-07-21T13:05:23.094195", "type": "content", "topic": "Building Multi-Agent Teams with LangGraph", "blog": "niteagent", "slug": "building-multi-agent-teams-with-langgraph", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.5, "word_count": 1588 }, { "model": "mimo-v2.5-pro", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.08, "word_count": 1650 }, { "model": "deepseek-v4", "rank": 3, "provider": "nexum", "family": "deepseek", "status": "failed", "error": "DialagramRouter upstream unavailable" } ], "judges": [ { "model": "llama-3.3-70b-versatile", "provider": "groq", "family": "llama", "role": "judge" }, { "model": "gpt-oss-120b", "provider": "cerebras", "family": "gpt", "role": "judge" }, { "model": "mistral-small-latest", "provider": "mistral", "family": "mistral", "role": "judge" } ], "winner_content_path": "/tmp/arena-outputs/guide-20260721-0300/deepseek-deepseek-v4-flash.mdx" }, { "match_id": "guide-20260722-1029", "date": "2026-07-22T10:29:55.910620", "type": "content", "topic": "Claude Code vs Cursor vs Windsurf: The Real-World 2026 Showdown", "blog": "niteagent", "slug": "claude-code-vs-cursor-vs-windsurf-2026-showdown", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.68, "word_count": 1658 }, { "model": "mimo-v2.5", "rank": 2, "provider": "custom", "family": "xiaomi", "status": "success", "score": 8.03, "word_count": 1500 } ], "judges": [ { "model": "llama-3.3-70b-versatile", "provider": "cerebras", "family": "llama", "role": "judge" }, { "model": "qwen3-32b", "provider": "groq", "family": "qwen", "role": "judge" }, { "model": "codestral-2508", "provider": "mistral", "family": "codestral", "role": "judge" }, { "model": "nemotron-3-ultra-550b", "provider": "openrouter", "family": "nemotron", "role": "judge" } ], "winner_content_path": "/home/techgeek/crypto-nite-blog/src/content/blog/claude-code-vs-cursor-vs-windsurf-2026-showdown.md" }, { "type": "app-clone", "date": "2026-07-23", "topic": "Match.com Clone — 4-step dating platform web app", "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "scores": { "css_compliance": 8.5, "functionality": 9.0, "code_quality": 7.5, "ux_design": 8.5 }, "avg_score": 8.55, "strengths": "Full localStorage persistence both steps, mobile chat UX with back button, search filtering, keyboard shortcuts, action badges on cards", "weaknesses": "Minified CSS hurts readability, innerHTML+= anti-pattern, short variable names", "score": 8.55 }, { "model": "mimo-v2.5", "provider": "xiaomi", "scores": { "css_compliance": 9.0, "functionality": 7.0, "code_quality": 8.5, "ux_design": 8.0 }, "avg_score": 8.1, "strengths": "Clean readable code, .map().join() rendering, red branded header, semantic HTML, aria-labels", "weaknesses": "NO localStorage in chat step — messages lost on refresh, no search wiring, limited mobile chat", "score": 8.1 } ], "winner": "deepseek-v4.1-flash", "winner_score": 8.55, "judges": [ { "model": "glm-5.2", "provider": "zai", "scores": { "a": 8.3, "b": 8.2 } }, { "model": "glm-5.2", "provider": "zai", "scores": { "a": 8.6, "b": 8.0 } }, { "model": "glm-5.2", "provider": "zai", "scores": { "a": 8.8, "b": 8.0 } }, { "model": "glm-5.2", "provider": "zai", "scores": { "a": 8.5, "b": 8.2 } } ] }, { "type": "app-clone", "date": "2026-07-23", "topic": "Match.com Clone v3 — arena rebuild with tutorial compliance", "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "scores": { "tutorial": 2, "css": 5.5, "functionality": 2.25, "code": 5 }, "avg_score": 3.5, "strengths": "Clean static HTML, responsive media queries", "weaknesses": "No JS, wrong header, missing all tutorial patterns", "score": 3.5 }, { "model": "mimo-v2.5", "provider": "xiaomi", "scores": { "tutorial": 8.25, "css": 8.5, "functionality": 8.75, "code": 8 }, "avg_score": 8.5, "strengths": "Full tutorial compliance, dark header, profile-first, all patterns present, full JS with localStorage", "weaknesses": "Logo still first in DOM order, data-score on article not span", "score": 8.5 } ], "winner": "mimo-v2.5", "winner_score": 8.5, "judges": [ { "model": "glm-5.2", "provider": "zai", "scores": { "a": 3, "b": 8 } }, { "model": "glm-5.2", "provider": "zai", "scores": { "a": 4, "b": 9 } }, { "model": "glm-5.2", "provider": "zai", "scores": { "a": 4, "b": 8 } }, { "model": "glm-5.2", "provider": "zai", "scores": { "a": 3, "b": 9 } } ] }, { "match_id": "daily-20260724-0001", "date": "2026-07-24T17:43:54.400147+00:00", "type": "content", "topic": "LLM Security Toolkit Stack: NeMo Guardrails vs LLM Guard vs Guardrails AI", "blog": "niteagent", "slug": "llm-security-toolkit-nemo-vs-llm-guard-vs-guardrails-ai", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 9.28, "word_count": 1497 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.75, "word_count": 1497 } ], "judges": [ { "model": "codestral-2508", "provider": "mistral", "family": "codestral", "score_a": 9.2, "score_b": 8.7, "winner": "A" }, { "model": "mistral-large-latest", "provider": "mistral", "family": "mistral", "score_a": 9.4, "score_b": 9.1, "winner": "A" }, { "model": "gemma-4-31b", "provider": "cerebras", "family": "gemma", "score_a": 9.1, "score_b": 8.4, "winner": "A" }, { "model": "mistral-medium-latest", "provider": "mistral", "family": "mistral", "score_a": 9.4, "score_b": 8.8, "winner": "A" } ], "winner_content_path": "/home/techgeek/crypto-nite-blog/src/content/blog/llm-security-toolkit-nemo-vs-llm-guard-vs-guardrails-ai.md" }, { "match_id": "daily-20260725-0001", "date": "2026-07-25T18:16:45.209034+00:00", "type": "content", "topic": "MCP 2026: Building Production MCP Servers — A Complete Developer Guide", "blog": "niteagent", "slug": "mcp-2026-building-production-mcp-servers-a-complete-developer-guide", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.98, "word_count": 1818 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.68, "word_count": 1348 } ], "judges": [ { "model": "codestral-2508", "provider": "mistral", "family": "codestral", "score_a": 8.2, "score_b": 8.5, "winner": "B" }, { "model": "mistral-medium-latest", "provider": "mistral", "family": "mistral", "score_a": 9.35, "score_b": 8.85, "winner": "A" }, { "model": "gemma-4-31b", "provider": "cerebras", "family": "gemma", "score_a": 9.4, "score_b": 8.7, "winner": "A" } ], "judge_note": "3 judges instead of 4: llama-3.3-70b-versatile (cerebras), qwen/qwen3-32b (groq), and nvidia/nemotron-3-ultra (openrouter) all returned 404 model_not_found; gpt-oss-120b (cerebras) hit an API format error; mistral-large-latest was rate-limited (429). Quorum of 3 achieved a clear verdict (A wins 2/3 judges, higher average).", "winner_content_path": "/home/techgeek/crypto-nite-blog/src/content/blog/2026-07-25-mcp-2026-building-production-mcp-servers-a-complete-developer-guide.md" }, { "match_id": "niteagent-roadmap-redesign-2026-07-31", "date": "2026-07-31", "type": "roadmap-redesign", "topic": "NiteAgent roadmap page redesign aligned to AI agent development", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "completed", "score": 8.5, "strengths": [ "15 linked posts", "AND-logic search+chips+progress", "no-JS fallback", "localStorage persistence" ], "weaknesses": [] }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "completed", "score": 7.0, "strengths": [ "sticky sidebar" ], "weaknesses": [ "only 10 posts", "less detailed interactions", "no reset" ] }, { "model": "qwen/qwen3.7-flash", "rank": 3, "provider": "openrouter", "family": "openrouter", "status": "completed", "score": 6.5, "strengths": [], "weaknesses": [ "spec cut off", "sessionStorage instead of localStorage", "<=10 posts" ] } ] }, { "match_id": "daily-20260804-0002", "date": "2026-08-04T16:00:00+00:00", "type": "content", "topic": "MCP Is Vertical, A2A Is Horizontal: A Protocol-Level Deep Dive into Agent-to-Agent Orchestration", "blog": "niteagent", "slug": "a2a-mcp-protocol-deep-dive", "winner": "xiaomi/mimo-v2.5", "results": [ { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.7, "word_count": 2657 }, { "model": "mimo-v2.5", "rank": 1, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 9.0, "word_count": 2545 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.2, "word_count": 2492 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "xiaomi", "score_a": 8.7, "score_b": 9.0, "score_c": 8.2, "verdict": "Winner: mimo-v2.5 (Draft B) with 9.0/10" } ] }, { "match_id": "daily-20260805-0001", "date": "2026-08-05T16:00:00+00:00", "type": "content", "topic": "MCP Transport Deep Dive: stdio vs Streamable HTTP — A Practical Decision Guide", "blog": "niteagent", "slug": "mcp-transport-deep-dive-stdio-vs-streamable-http", "winner": "deepseek/deepseek-v4-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 9.4, "word_count": 2296 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "xiaomi", "status": "success", "score": 8.8, "word_count": 1993 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 9.1, "word_count": 2255 } ], "judges": [ { "model": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score_a": 9.4, "score_b": 8.8, "score_c": 9.1, "verdict": "Winner: deepseek/deepseek-v4-flash (Draft A) with 9.4/10" } ] }, { "match_id": "daily-20260806-0001", "date": "2026-08-06T06:00:00+00:00", "type": "content", "topic": "AI Is Hacking Crypto Faster Than Humans Can Fix It", "blog": "niteagent", "slug": "2026-08-06-ai-hacking-crypto-faster-than-humans-can-fix-it", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.43, "word_count": 2333, "sources": 11 }, { "model": "deepseek-v4.1-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.14, "word_count": 1983, "sources": 9 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.14, "word_count": 1983, "sources": 9 } ], "judges": [ { "model": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.43, "status": "pass", "role": "judge" } ] }, { "match_id": "tool-match-clone-rebuild-2026-08-06", "date": "2026-08-06T15:00:00+00:00", "type": "app", "topic": "Match.com Clone — Modern Redesign (mobile-first swipe-deck dating app)", "blog": "niteagent", "slug": "apps/match-clone/step4", "winner": "poolside-laguna-s-2.1", "threshold_met": true, "results": [ { "model": "poolside-laguna-s-2.1", "rank": 1, "provider": "openrouter", "family": "poolside", "status": "success", "score": 9.72 }, { "model": "cmd-codex", "rank": 2, "provider": "cmd", "family": "cmd", "status": "success", "score": 9.23 }, { "model": "deepseek-v4.1-flash", "rank": 3, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 9.15 }, { "model": "mimo-v2.5", "rank": 4, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 8.93 } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "score": 9.72, "status": "pass", "role": "judge" } ] }, { "match_id": "mcp-stateless-rewrite-2026-08-06", "date": "2026-08-06T16:00:00+00:00", "type": "deep-dive", "topic": "MCP Goes Stateless: Inside the 2026-07-28 Spec Rewrite", "blog": "niteagent", "slug": "2026-08-06-mcp-goes-stateless-2026-07-28-spec-rewrite", "winner": "deepseek-v4.1-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 9.0 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 8.3 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.3 } ], "judges": [ { "model": "mimo-v2.5-pro", "provider": "xiaomi", "family": "mimo", "status": "pass", "role": "judge" } ] }, { "match_id": "mcp-oauth21-resource-server-field-guide-2026", "date": "2026-08-09", "type": "content", "topic": "MCP OAuth 2.1 resource server authorization implementation guide", "blog": "niteagent", "slug": "mcp-oauth21-resource-server-field-guide-2026", "winner": "deepseek-v4.1-flash", "results": [ { "rank": 1, "model": "deepseek-v4.1-flash", "provider": "deepseek", "family": "deepseek", "score": 8.5, "word_count": 1331, "status": "pass" }, { "rank": 2, "model": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 7.2, "word_count": 1304, "status": "pass" }, { "rank": 3, "model": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 0.0, "word_count": 0, "status": "fail", "error": "Circuit breaker open for openrouter" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "role": "judge", "score": 8.5 }, { "model": "glm-5.2", "provider": "zai", "role": "qa_gate", "score": 8.5, "pass": true } ] }, { "match_id": "arena-mast-failures", "date": "2026-08-11", "type": "content", "topic": "Why Do Multi-Agent Systems Fail? The MAST Taxonomy, Decoded", "blog": "niteagent", "slug": "why-multi-agent-systems-fail-mast-taxonomy", "winner": "deepseek-v4.1-flash", "results": [ { "rank": 1, "model": "deepseek-v4.1-flash", "provider": "deepseek", "family": "deepseek", "score": 8.6, "word_count": 1674, "status": "pass" }, { "rank": 2, "model": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 6.4, "word_count": 1211, "status": "pass" }, { "rank": 3, "model": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 0.0, "word_count": 0, "status": "fail", "error": "API error: openrouter/poolside/laguna-s-2.1 is not a valid model ID" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "role": "judge", "score": 8.6 }, { "model": "glm-5.2", "provider": "zai", "role": "qa_gate", "score": 6.6, "pass": false, "note": "QA failed on original draft (wrong taxonomy names, fabricated code API). Draft corrected with verified paper data and re-shipped." } ] }, { "match_id": "arena-na-widescreen-redesign-2026-08-11", "date": "2026-08-11", "type": "redesign", "topic": "NiteAgent Arena page — widescreen + space optimization redesign", "blog": "niteagent", "winner": "poolside-laguna-s-2.1", "results": [ { "rank": 1, "model": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.15, "word_count": 2764, "status": "pass" }, { "rank": 2, "model": "deepseek-v4.1-flash", "provider": "deepseek", "family": "deepseek", "score": 7.9, "word_count": 2910, "status": "pass" }, { "rank": 3, "model": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 6.33, "word_count": 2736, "status": "pass" } ], "judges": [ { "model": "qwen3.7-flash", "winner": "poolside/laguna-s-2.1" } ], "judge_passes": 3, "threshold_met": true }, { "match_id": "arena-nvidia-depin-gap-2026-08-12", "date": "2026-08-12", "type": "content", "topic": "Nvidia's $500B Wall Street Compute Play: What It Means for Decentralized AI Tokens", "blog": "niteagent", "slug": "nvidia-500b-depin-compute-gap", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.0, "word_count": 1326 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 7.0, "word_count": 1170 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "failed", "score": 0.0, "word_count": 0, "error": "API error: openrouter/poolside/laguna-s-2.1 is not a valid model ID" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "score_a": 8.0, "score_b": 7.0, "winner": "A" }, { "model": "glm-5.2", "provider": "zai", "family": "glm", "qa_score": 8.5, "qa_pass": true } ] }, { "match_id": "arena-mcp-stateless-2026", "date": "2026-08-12", "type": "content", "topic": "MCP Went Stateless: Inside the July 28, 2026 Protocol Revision", "blog": "niteagent", "slug": "mcp-went-stateless-july-2026", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 4.5, "word_count": 1122, "rank": 1 }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 4.5, "word_count": 1130, "rank": 2 }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 0.0, "word_count": 0, "rank": 3, "error": "API error: invalid model ID" } ], "judges": [ { "model": "deepseek-v4-flash", "provider": "deepseek", "role": "judge", "draft_a_score": 5.7, "draft_b_score": 4.0, "winner": "a" }, { "model": "deepseek-v4-flash", "provider": "deepseek", "role": "qa_gate", "initial_score": 7.8, "final_score": 8.5, "pass": true, "fixes_applied": [ "Namespaced _meta keys", "cacheScope global -> public", "stateToken -> requestState", "inputSchema -> inputRequests", "Removed alphabetical ordering claim", "Fixed Python SDK import path", "notifications/initialized", "Fixed citation attribution", "Sentence case headings" ] } ] }, { "match_id": "arena-bitcoin-miners-ai-pivot-aug2026", "date": "2026-08-13", "type": "content", "topic": "Bitcoin Miners Are Becoming AI Companies — Selling BTC to Fund the AI Compute Race", "blog": "niteagent", "slug": "bitcoin-miners-ai-pivot-aug2026", "winner": "deepseek-v4.1-flash", "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 4.5, "word_count": 1231 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 3.5, "word_count": 1489 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "failed", "score": 0.0, "word_count": 0, "error": "API error: openrouter/poolside/laguna-s-2.1 is not a valid model ID" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "score_a": 8.5, "score_b": 7.0, "winner": "A" }, { "model": "glm-5.2", "provider": "zai", "family": "glm", "qa_score": 8.5, "qa_pass": true, "note": "Citation URLs fixed after initial QA fail (hallucinated slugs)" } ] }, { "match_id": "arena-verify-ai-inference-attestation-zk-guide", "date": "2026-08-14", "type": "content", "topic": "Don't Trust the Model, Verify It: A Practical Guide to Verifiable AI Inference (Attestation + ZK Proofs)", "blog": "niteagent", "slug": "verify-ai-inference-attestation-zk-guide", "content_type": "guide", "plan_model": "deepseek-v4-flash", "winner": "deepseek-v4.1-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 9.77, "word_count": 2197, "per_pass_scores": [ 9.9, 9.7, 9.7 ], "mean_score": 9.77 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 9.13, "word_count": 2474, "per_pass_scores": [ 9.7, 9.0, 8.7 ], "mean_score": 9.13 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "failed", "score": 0.0, "word_count": 0, "error": "Timeout (60s)" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "passes": 3, "mean_weighted": 9.77, "rubric": "accuracy 30% | sourcing 25% | quality 20% | practicality 15% | engagement 10%", "verified_fact_preamble": true } ], "qa": { "model": "deepseek-v4-flash", "provider": "deepseek", "note": "cron delegation pinned to deepseek-v4-flash (glm-5.2 not routable via delegate_task)", "verdict": "ready to ship after 1 P0 fix", "p0_fixed": "frontmatter desc -> description (schema requires description: z.string())", "fabrication_gate": "pass", "canonical_honesty_sentence": "present verbatim", "link_check": "all external + internal links HTTP 200" }, "notes": "Builder B mimo-v2.5, builder D poolside timed out (DNF). Planner + QA ran deepseek-v4-flash via delegate_task (cron delegation config pins subagents; mimo-v2.5-pro/glm-5.2 not routable). Research ran deepseek-v4-flash via delegate_task (gpt-5.6-luna not routable). Judge qwen3.7-flash multipass 3x, mean-gated." }, { "match_id": "arena-mcp-server-performance-benchmarking-2026", "date": "2026-08-14", "type": "content", "topic": "MCP Server Performance Benchmarking: What to Measure, What the Numbers Say, and What Changed in the 2026-07-28 Spec", "blog": "niteagent", "slug": "mcp-server-performance-benchmarking-2026", "content_type": "benchmarks", "plan_model": "deepseek-v4-flash", "winner": "deepseek-v4.1-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.9, "word_count": 2504, "per_pass_scores": [ 9.5, 8.5, 8.7 ], "mean_score": 8.9 }, { "model": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 8.67, "word_count": 1420, "per_pass_scores": [ 9.05, 8.1, 8.85 ], "mean_score": 8.67 }, { "model": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "failed", "score": 0.0, "word_count": 0, "per_pass_scores": null, "mean_score": null, "error": "Timeout (60s)" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "passes": 3, "mean_weighted": 8.9 } ] }, { "match_id": "arena-anthropic-multi-agent-red-team-trading-risks", "date": "2026-08-15", "type": "news-analysis", "topic": "Anthropic's Red Team Found Multi-Agent Sabotage and Collusion — What It Means for Crypto Trading Bots", "blog": "niteagent", "slug": "anthropic-multi-agent-red-team-trading-risks", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "status": "success", "word_count": 2009, "judge_scores": [ 7.8, 7.9, 8.05 ], "judge_mean": 7.92, "score": 7.92 }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "word_count": 1635, "judge_scores": [ 8.9, 8.5, 8.9 ], "judge_mean": 8.77, "score": 8.77 }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "word_count": 1881, "judge_scores": [ 8.2, 8.1, 7.1 ], "judge_mean": 7.8, "score": 7.8 } ] }, { "match_id": "arena-agent-sandboxing-2026", "date": "2026-08-15", "type": "content", "topic": "Production Sandboxing for AI Agents: Containment Patterns for Untrusted Tool Execution and MCP Servers", "blog": "niteagent", "slug": "2026-08-15-agent-sandboxing-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "status": "success", "word_count": 3430, "judge_scores": [ 8.3, 9.45, 8.7 ], "judge_mean": 8.82, "score": 8.82 }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "word_count": 3217, "judge_scores": [ 8.2, 8.6, 9.05 ], "judge_mean": 8.62, "score": 8.62 }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "word_count": 2443, "judge_scores": [ 8.1, 8.75, 8.7 ], "judge_mean": 8.52, "score": 8.52 } ] }, { "match_id": "arena-cloudflare-wallets-x402-2026", "date": "2026-08-16", "type": "content", "topic": "The Agentic Payments Stack: x402 + Cloudflare Wallets — how AI agents got wallets, identity, and pay-per-request economics (paper anchor: x402 whitepaper Jun 2026; news anchor: Cloudflare Wallets Aug 4 2026)", "blog": "niteagent", "slug": "cloudflare-wallets-x402-agentic-payments-2026", "winner": "mimo-v2.5", "plan_model": "deepseek-v4-flash", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "status": "success", "word_count": 1444, "judge_scores": [ 8.2, 8.95, 8.75 ], "judge_mean": 8.63, "score": 8.63, "judge_model": "openrouter/qwen3.7-flash" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "word_count": 1601, "judge_scores": [ 8.75, 8.85, 8.65 ], "judge_mean": 8.75, "score": 8.75, "judge_model": "openrouter/qwen3.7-flash" }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "word_count": 2504, "judge_scores": [ 9.15, 8.55, 7.4 ], "judge_mean": 8.37, "score": 8.37, "judge_model": "openrouter/qwen3.7-flash" } ] }, { "match_id": "arena-ai-compute-futures-gpu-cost-guide", "date": "2026-08-17", "type": "content", "topic": "Compute futures are coming: how AI engineers should benchmark, budget, and hedge GPU costs (CME + Silicon Data H100/B200 rental-index futures, Aug 11 2026; CFTC review began Aug 17 2026; Python GPU-cost intelligence pipeline + worked hedging decision)", "blog": "niteagent", "slug": "ai-compute-futures-gpu-cost-guide", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "status": "success", "word_count": 3124, "judge_scores": [ 9.05, 9.5, 9.05 ], "judge_mean": 9.2, "score": 9.2, "judge_model": "openrouter/qwen3.7-flash" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "judge_scores": [ 9.05, 9.15, 9.05 ], "judge_mean": 9.08, "score": 9.08, "judge_model": "openrouter/qwen3.7-flash" }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "judge_scores": [ 8.8, 8.5, 8.5 ], "judge_mean": 8.6, "score": 8.6, "judge_model": "openrouter/qwen3.7-flash" } ] }, { "match_id": "arena-prompt-injection-defense-production-2026", "date": "2026-08-18", "type": "content", "topic": "Prompt Injection Defense in Production: detecting, containing, and hardening AI agents against untrusted input (direct/indirect injection, tool poisoning, detection classifiers, containment layers, output verification, red-teaming) — layered lifecycle playbook", "blog": "niteagent", "slug": "prompt-injection-defense-production-2026", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "status": "success", "word_count": 1926, "judge_scores": [ 8.95, 8.45, 8.05 ], "judge_mean": 8.48, "score": 8.48, "judge_model": "openrouter/qwen3.7-flash" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "word_count": 1906, "judge_scores": [ 8.95, 8.95, 9.2 ], "judge_mean": 9.03, "score": 9.03, "judge_model": "openrouter/qwen3.7-flash" }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "word_count": 2024, "judge_scores": [ 8.45, 8.95, 9.6 ], "judge_mean": 9.0, "score": 9.0, "judge_model": "openrouter/qwen3.7-flash" } ] }, { "match_id": "arena-ai-agent-tokens-crash-2026", "date": "2026-08-19", "type": "benchmark", "topic": "AI Agent Tokens Crash 2026: ELIZAOS's $2.5B Collapse vs FLOP's Fair Launch Reset vs VIRTUAL and FET", "blog": "niteagent", "slug": "ai-agent-tokens-crash-2026", "winner": "deepseek-v4.1-flash", "plan_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "rank": 1, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.95, "per_pass": [ 9.55, 8.55, 8.75 ], "word_count": 2640, "judge_note": "qwen3.7-flash primary; deepseek-v4-flash-0731 fallback on 429 passes" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 8.57, "per_pass": [ 9.55, 8.1, 8.05 ], "word_count": 1809 }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 4.27, "per_pass": [ 5.05, 3.95, 3.8 ], "word_count": 2854, "note": "no inline markdown links -> sourcing 1.0" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "method": "tb-judge-multipass.py --passes 3, mean-gated", "note": "mean 8.95/8.57/4.27; qwen 429s on some passes used deepseek/deepseek-v4-flash-0731 fallback" } ] }, { "match_id": "arena-multi-tenant-agent-platforms-2026", "date": "2026-08-19", "type": "content", "topic": "Multi-Tenant AI Agent Platforms: Isolation Patterns for the LLM Era — pool/silo/bridge patterns, RLS vs schema-per-tenant vs database-per-tenant, per-tenant token budgets, MCP credential safety, tenant-aware observability", "blog": "niteagent", "slug": "multi-tenant-agent-platform-architecture-2026", "winner": "poolside-laguna-s-2.1", "plan_model": "mimo-v2.5-pro", "threshold_met": true, "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "notes": "Winner condensed from 6098w draft to ~3750w (site norms) post-judge; 30 source URLs verified HTTP 200; QA (GLM-5.2) issues fixed (Temporal overclaim, internal-link density, dedup); heroImage via gen-image-verified.sh", "results": [ { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "rank": 1, "provider": "openrouter", "family": "poolside", "status": "success", "score": 9.17, "per_pass": [ 9.15, 9.4, 8.95 ], "word_count": 6098, "judge_note": "winner; quality 7-9 across passes; length over 2500 target but not penalized by rubric; condensed at finalize" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "rank": 2, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 9.12, "per_pass": [ 9.05, 9.15, 9.15 ], "word_count": 2767 }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "rank": 3, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.87, "per_pass": [ 9.4, 8.85, 8.35 ], "word_count": 3621 } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "method": "tb-judge-multipass.py --passes 3, mean-gated", "note": "means 9.17/9.12/8.87, all >= 8.0 PASS; verified-fact preamble prepended for judging" } ] }, { "match_id": "arena-binance-mcp-server-guardrailed-ai-trading-agent", "date": "2026-08-20", "type": "content", "topic": "I Gave Claude Code a Live Binance Account: Building a Guardrailed AI Trading Agent with the Binance MCP Server — Agent OS launch (2026-08-20), no-withdrawal scope, confirm-before-execute, Agentic sub-account isolation, least-privilege OAuth scopes, MCP initialize handshake (HTTP 200)", "blog": "niteagent", "slug": "binance-mcp-server-guardrailed-ai-trading-agent", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "threshold_met": true, "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "notes": "Build Log (Thu). Verified-fact preamble prepended for judging; all 3 builders PASS >= 8.0; winner mimo-v2.5 (8.95); heroImage via gen-image-verified.sh (Lumenfall FLUX, HTTP 200); QA live-fact-checked all external links (binance.com 202 anti-bot accepted, content extracted; theblock 403 excluded); no fabricated hands-on trading claims (canonical E-E-A-T phrasing used)", "results": [ { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "rank": 1, "provider": "xiaomi", "family": "mimo", "status": "success", "score": 8.95, "per_pass": [ 8.95, 9.15, 8.75 ], "word_count": 2147, "judge_note": "winner; accuracy 9-10, sourcing 10 across passes" }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "rank": 2, "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.4, "per_pass": [ 8.55, 8.8, 7.85 ], "word_count": 2596 }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "rank": 3, "provider": "openrouter", "family": "poolside", "status": "success", "score": 8.32, "per_pass": [ 9.2, 8.8, 6.95 ], "word_count": 2283, "judge_note": "pass spread 2.25 > 1.5 (noisy judge) but mean >= 8.0" } ], "judges": [ { "model": "qwen3.7-flash", "provider": "openrouter", "family": "qwen", "method": "tb-judge-multipass.py --passes 3, mean-gated", "note": "means 8.95/8.40/8.32, all >= 8.0 PASS; verified-fact preamble prepended for judging" } ] }, { "match_id": "arena-vllm-vs-sglang-vs-tensorrt-llm-2026", "date": "2026-08-20", "type": "content", "topic": "vLLM vs SGLang vs TensorRT-LLM in 2026: LLM inference engine benchmark deep-dive", "blog": "niteagent", "slug": "vllm-vs-sglang-vs-tensorrt-llm-2026", "winner": "deepseek-v4.1-flash", "plan_model": "deepseek-v4-flash", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "family": "deepseek", "score": 9.28, "status": "success", "word_count": 2692, "scores": { "pass1": 9.6, "pass2": 9.5, "pass3": 8.75 }, "per_pass": [ 9.6, 9.5, 8.75 ], "strengths": [ "Thorough (2692 words)" ], "model_id": "deepseek-v4-flash" }, { "model": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.17, "status": "success", "word_count": 2381, "scores": { "pass1": 8.15, "pass2": 8.5, "pass3": 7.85 }, "per_pass": [ 8.15, 8.5, 7.85 ], "strengths": [ "Thorough (2381 words)" ], "model_id": "mimo-v2.5" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 7.52, "status": "success", "word_count": 2088, "scores": { "pass1": 8.0, "pass2": 7.25, "pass3": 7.3 }, "per_pass": [ 8.0, 7.25, 7.3 ], "strengths": [ "Thorough (2088 words)" ], "weaknesses": [ "Sourcing criterion weak in passes 2-3 (4.0-5.0)" ], "model_id": "poolside-laguna-s-2.1" } ] }, { "match_id": "arena-agentic-payments-security-guide-2026", "date": "2026-08-21", "type": "guide-tutorial", "topic": "Agentic Payments Security Guide 2026 — x402 hit ~14M agent transfers in 30 days; security-first field guide for shipping AI agents with wallets (payment flow, revenue-attached tokens, 7-point hardening checklist)", "blog": "niteagent", "slug": "agentic-payments-security-guide-2026", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 8.18, "status": "success", "per_pass": [ 8.6, 8.0, 7.95 ], "judge_mean": 8.18, "word_count": 2436, "strengths": [ "Strong sourcing, complete structure" ], "weaknesses": [], "model_id": "deepseek-v4-flash" }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.43, "status": "success", "per_pass": [ 8.2, 8.4, 8.7 ], "judge_mean": 8.43, "word_count": 1987, "strengths": [ "Best accuracy + sourcing, clean structure" ], "weaknesses": [], "model_id": "mimo-v2.5" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 0.0, "status": "dnf", "error": "Degenerate output: echoed the input plan verbatim (228 lines, plan headings only) — not a blog post. Recorded as DNF; excluded from leaderboard.", "word_count": 2981, "model_id": "poolside-laguna-s-2.1" } ] }, { "match_id": "arena-text-to-sql-agents-production-2026", "date": "2026-08-21", "type": "content", "topic": "Text-to-SQL Agents in Production 2026: Architectures, Safety, and Evaluation", "blog": "niteagent", "slug": "text-to-sql-agents-production-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "openai/gpt-5.6-luna", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "family": "deepseek", "score": 8.62, "status": "success", "word_count": 2768, "scores": { "pass1": 8.35, "pass2": 8.65, "pass3": 8.85 }, "per_pass": [ 8.35, 8.65, 8.85 ], "strengths": [ "Strongest sourcing (12 unique external links)", "Complete outline coverage", "Clean E-E-A-T block" ], "model_id": "deepseek-v4-flash" }, { "model": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.43, "status": "success", "word_count": 1639, "scores": { "pass1": 8.4, "pass2": 8.5, "pass3": 8.4 }, "per_pass": [ 8.4, 8.5, 8.4 ], "strengths": [ "Consistent scoring across passes" ], "model_id": "mimo-v2.5" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.37, "status": "success", "word_count": 3585, "scores": { "pass1": 8.4, "pass2": 8.0, "pass3": 8.7 }, "per_pass": [ 8.4, 8.0, 8.7 ], "strengths": [ "Longest draft (3585 words)" ], "weaknesses": [ "Quality criterion dipped in pass 2" ], "model_id": "poolside-laguna-s-2.1" } ] }, { "match_id": "arena-sec-regulation-crypto-assets-ai-tokens", "date": "2026-08-22", "type": "news-analysis", "topic": "SEC Regulation Crypto Assets proposal: delinking safe harbor and new fundraising exemptions — what they mean for AI tokens", "blog": "niteagent", "slug": "sec-regulation-crypto-assets-ai-tokens", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 8.3, "scores": { "pass1": 9.15, "pass2": 8.8, "pass3": 6.95, "mean": 8.3 }, "word_count": 2836, "strengths": [ "Thorough coverage of delinking mechanism" ], "weaknesses": [ "Quality variance across passes" ], "model_id": "deepseek-v4-flash" }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 9.03, "scores": { "pass1": 8.6, "pass2": 9.15, "pass3": 9.35, "mean": 9.03 }, "word_count": 2126, "strengths": [ "High accuracy and sourcing across all passes", "Best overall structure" ], "weaknesses": [], "model_id": "mimo-v2.5" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 8.28, "scores": { "pass1": 7.85, "pass2": 8.2, "pass3": 8.8, "mean": 8.28 }, "word_count": 3062, "strengths": [ "Comprehensive, long-form coverage" ], "weaknesses": [ "Quality criterion low on first passes" ], "model_id": "poolside-laguna-s-2.1" } ] }, { "match_id": "arena-mlperf-client-v2-agentic-ai-2026", "date": "2026-08-22", "type": "content", "topic": "MLPerf Client v2.0: The First Agentic AI Benchmark for AI PCs — what the new agentic/image workloads measure and what v1.0-era vendor numbers mean for shipping local agents", "blog": "niteagent", "slug": "mlperf-client-v2-agentic-ai-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "per_pass": [ 8.65, 9.5, 9.7 ], "judge_mean": 9.28, "status": "fail", "score": 9.28 }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "per_pass": [ 9.15, 8.85, 9.55 ], "judge_mean": 9.18, "status": "fail", "score": 9.18 }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "per_pass": [ 7.65, 9.35, 8.85 ], "judge_mean": 8.62, "status": "fail", "score": 8.62 } ] }, { "match_id": "arena-when-agents-act-on-web3-mcp-attack-surfaces", "date": "2026-08-23", "type": "content", "topic": "When Agents Act on Web3: The MCP Attack Surface Is Worse Than You Think (arXiv:2608.17275 paper explainer + Chainlink for Agents / Gensyn Delphi Arena / Binance Agent OS)", "blog": "niteagent", "slug": "when-agents-act-on-web3-mcp-attack-surfaces", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 9.07, "word_count": 1791, "status": "success", "scores": { "pass1": 8.9, "pass2": 8.95, "pass3": 9.35, "mean": 9.07 }, "strengths": [ "All 27%->65% / <30% / <3% stats inline-sourced to arXiv", "2 comparison tables", "6 internal cluster links" ], "weaknesses": [ "Some opening H2 paragraphs above the 40-60 word target" ] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.9, "word_count": 1695, "status": "success", "scores": { "pass1": 9.35, "pass2": 9.1, "pass3": 8.25, "mean": 8.9 }, "strengths": [ "Clean inline sourcing", "Solid structure" ], "weaknesses": [ "Word count near floor" ] }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 7.77, "word_count": 2479, "status": "success", "scores": { "pass1": 7.5, "pass2": 7.3, "pass3": 8.5, "mean": 7.77 }, "strengths": [ "Longest draft (2479 words)" ], "weaknesses": [ "Quality sub-8.0 across passes", "Did not win" ] } ] }, { "match_id": "arena-agentic-rag-architecture-2026", "date": "2026-08-23", "type": "content", "topic": "Agentic RAG Architecture: The Four Control Points That Make Retrieval Production-Ready (research lineage Self-RAG/CRAG/Adaptive-RAG/DeepRAG/ITER-RETGEN/GraphRAG mapped to decompose-route-retrieve-verify production control points)", "blog": "niteagent", "slug": "agentic-rag-architecture-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.53, "word_count": 1803, "status": "success", "scores": { "pass1": 8.15, "pass2": 9.1, "pass3": 8.35, "mean": 8.53 }, "strengths": [ "Inline-sourced stats (DeepRAG +26.4%, RouteLLM >2x)", "10 H2 sections", "5 FAQ answers self-contained", "Comparison table 5x5" ], "weaknesses": [ "QA edits applied post-judge: H2 openers trimmed to 40-60 words, word count 1765 shipped" ] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.48, "word_count": 1088, "status": "success", "scores": { "pass1": 8.6, "pass2": 7.85, "pass3": 9.0, "mean": 8.48 }, "strengths": [ "Clean structure", "Strong per-section sourcing" ], "weaknesses": [ "Word count 1088 — below 1500 target" ] }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.18, "word_count": 2680, "status": "success", "scores": { "pass1": 8.35, "pass2": 8.3, "pass3": 7.9, "mean": 8.18 }, "strengths": [ "Longest draft (2680 words)" ], "weaknesses": [ "Quality variance across passes", "Lost to deepseek/mimo on mean" ] } ] }, { "match_id": "arena-virtuals-solana-agent-token-ownership-guide", "date": "2026-08-24", "type": "content", "topic": "AI Agent Ownership on Solana: How Virtuals' EconomyOS Works and How to Evaluate Agent Tokens After the ai16z Collapse", "blog": "niteagent", "slug": "virtuals-solana-agent-token-ownership-guide", "winner": "poolside-laguna-s-2.1", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 9.15, "per_pass": [ 8.75, 9.15, 9.55 ], "word_count": 2155, "status": "success", "note": "WINNER — highest multipass mean; judge qwen3.7-flash" }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.65, "per_pass": [ 8.55, 9.15, 8.25 ], "word_count": 2260, "status": "success", "note": "judge qwen3.7-flash; pass 2 fell back to deepseek/deepseek-v4-flash-0731 via OpenRouter on 429" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 7.23, "per_pass": [ 6.95, 7.65, 7.1 ], "word_count": 1824, "status": "success", "note": "below 8.0 threshold — weak sourcing (2-4 on 25% criterion)" } ] }, { "match_id": "arena-durable-execution-agents-2026", "date": "2026-08-24", "type": "content", "topic": "Checkpointing is not durable execution: keeping long-running AI agents alive (Temporal/DBOS/Restate durable-execution engines vs LangGraph/OpenAI Agents SDK checkpointers — naive-retry failure taxonomy, saga/idempotency/resume patterns, decision framework)", "blog": "niteagent", "slug": "durable-execution-agents-2026", "winner": "poolside-laguna-s-2.1", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 9.17, "word_count": 3953, "status": "success", "scores": { "pass1": 9.7, "pass2": 8.1, "pass3": 9.7, "mean": 9.17 }, "strengths": [ "Highest accuracy/sourcing per-pass (10/10 on both in 2 of 3 passes)", "Clean structure with 10 H2s", "Strong code sketch" ], "weaknesses": [ "Quality variance pass2 8.10 (noise)", "Slightly less table depth than poolside" ] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.88, "word_count": 3690, "status": "success", "scores": { "pass1": 9.15, "pass2": 9.15, "pass3": 8.35, "mean": 8.88 }, "strengths": [ "Very consistent sourcing (10/10 all passes)", "Solid accuracy", "Good FAQ depth" ], "weaknesses": [ "Third-place mean", "Quality dipped to 5.0 in pass 3" ] }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 9.63, "word_count": 4359, "status": "success", "scores": { "pass1": 9.8, "pass2": 9.2, "pass3": 9.9, "mean": 9.63 }, "strengths": [ "Winner — highest mean 9.63", "10/10 accuracy and sourcing", "2 comparison tables", "37 sections (10 H2 + 27 H3)", "QA-verified: 25 unique external URLs all 200 (openai.com 403-to-curl allowed), 77 link instances, zero slop, zero fabrication" ], "weaknesses": [ "Prose ~3027 words above 2500 ideal (advisory)", "Two minor citations fixed by QA (Restate BSL re-cite, Inngest quote verbatim)" ] } ] }, { "match_id": "arena-agentic-payments-settlement-stack-2026", "date": "2026-08-25", "type": "architecture-deep-dive", "topic": "The agentic payments settlement stack: x402 + USDC + Circle Arc L1 (99.3% USDC settlement, Arc mainnet Sept 16, x402 Foundation)", "blog": "niteagent", "slug": "agentic-payments-settlement-stack-2026", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "family": "deepseek", "status": "pass", "per_pass_scores": [ 7.55, 9.05, 9.45 ], "mean": 8.68, "word_count": 2919, "model_id": "deepseek-v4-flash", "score": 8.68 }, { "model": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "pass", "per_pass_scores": [ 9.45, 9.05, 8.95 ], "mean": 9.15, "word_count": 2863, "winner": true, "model_id": "mimo-v2.5", "score": 9.15 }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "pass", "per_pass_scores": [ 8.3, 9.35, 8.05 ], "mean": 8.57, "word_count": 3128, "model_id": "poolside-laguna-s-2.1", "score": 8.57 } ] }, { "match_id": "arena-semantic-llm-routing-production", "date": "2026-08-25", "type": "content", "topic": "The Router Layer: Semantic & Cost-Aware LLM Routing in Production", "blog": "niteagent", "slug": "semantic-llm-routing-production", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.8, "scores": [ 9.55, 8.0, 8.85 ], "status": "success", "word_count": 2632, "note": "pass 1 judge fell back to deepseek/deepseek-v4-flash-0731 after qwen 429; passes 2-3 qwen" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 9.13, "scores": [ 8.85, 8.85, 9.7 ], "status": "success", "word_count": 1931, "note": "winner after QA fill (FrugalGPT citation, +1 internal link)" }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.55, "scores": [ 8.45, 8.7, 8.5 ], "status": "success", "word_count": 3784 } ] }, { "match_id": "arena-cryptobench-leaderboard-ai-crypto-agents-2026", "date": "2026-08-26", "type": "benchmark-comparison", "topic": "CryptoBench leaderboard deep-dive: which LLM agents actually win at crypto-native analysis, and what the retrieval-prediction gap means for the new agent-execution platforms (Binance Agent OS, Coinbase for Agents, Rain Agentic Payments Alliance)", "blog": "niteagent", "slug": "cryptobench-leaderboard-ai-crypto-agents-2026", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "status": "success", "rank": 3, "score": 8.13, "word_count": 2163, "per_pass_scores": [ 7.9, 8.85, 7.65 ], "judge_model": "qwen3.7-flash" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "rank": 1, "score": 9.08, "word_count": 2316, "per_pass_scores": [ 8.55, 9.35, 9.35 ], "judge_model": "qwen3.7-flash", "qa_fixes": "corrected 44 fabricated benchmark scores against arXiv:2512.00417v5 Tables 3-4; added missing Table B (BeInCrypto product scores) with editorial caveat; expanded FAQ answer 3; added 3 internal links" }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "rank": 2, "score": 8.2, "word_count": 1643, "per_pass_scores": [ 6.85, 9.6, 8.15 ], "judge_model": "qwen3.7-flash" } ] }, { "match_id": "arena-mcp-stateless-2026-spec-deep-dive", "date": "2026-08-26", "type": "content", "topic": "MCP Goes Stateless: Inside the 2026-07-28 Spec — and How to Migrate Your Production Servers (stateless protocol core, MRTR, Mcp-Method/Mcp-Name header routing, cacheable tool catalogs, Extensions framework, OAuth 2.1 hardening, Python/TypeScript SDK v2 migration)", "blog": "niteagent", "slug": "mcp-stateless-2026-spec-deep-dive", "winner": "poolside-laguna-s-2.1", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.6, "word_count": 3477, "status": "success", "scores": { "pass1": 8.85, "pass2": 7.9, "pass3": 9.05, "mean": 8.6 }, "strengths": [ "Solid accuracy (9/9/9) and sourcing (10/9/10) across passes" ], "weaknesses": [ "Quality variance: pass2 quality dropped to 4.0" ] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.95, "word_count": 2972, "status": "success", "scores": { "pass1": 9.7, "pass2": 8.5, "pass3": 8.65, "mean": 8.95 }, "strengths": [ "Highest single-pass accuracy/sourcing (10/10 in pass1 and pass3)", "Clean structure" ], "weaknesses": [ "Quality dipped to 5.0 in pass3 (judge noise)" ] }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 9.17, "word_count": 2508, "status": "success", "scores": { "pass1": 8.5, "pass2": 9.3, "pass3": 9.7, "mean": 9.17 }, "strengths": [ "WINNER — highest mean 9.17", "10/10 accuracy+sourcing in 2 of 3 passes", "2 comparison tables", "FAQ 6 Q&As", "QA-verified: 19 unique external URLs all HTTP 200, no fabrication, canonical E-E-A-T block" ], "weaknesses": [ "Pass1 quality 7.0 (noise)" ] } ] }, { "match_id": "arena-ai-agents-tokenized-equities-2026", "date": "2026-08-27", "type": "build-log", "topic": "AI agents get equities: the Coinbase B20 tokenized-stock rail on Base (issuance) + Chainlink multiplier feeds (pricing) + Alpaca/ADGM custody + Bitwise ATPs (rebalancing) + Virtuals agent wallets — Aug 24-27, 2026", "blog": "niteagent", "slug": "ai-agents-tokenized-equities-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.52, "scores": { "passes": [ 8.3, 8.55, 8.7 ], "mean": 8.52 }, "word_count": 2438, "strengths": [ "Accurate sourcing", "Strong inline links", "Honest E-E-A-T disclosure", "13 H2 sections, FAQ, comparison table" ], "weaknesses": [] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 7.73, "scores": { "passes": [ 7.6, 7.6, 8.0 ], "mean": 7.73 }, "word_count": 1757, "strengths": [ "Accurate facts", "Good sourcing" ], "weaknesses": [ "Quality sub-8 (structure/self-contained answers)" ] }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 7.17, "scores": { "passes": [ 6.0, 6.7, 8.8 ], "mean": 7.17 }, "word_count": 1893, "strengths": [ "Thorough" ], "weaknesses": [ "Noisy passes (spread 2.8), sourcing inconsistent across passes" ] } ] }, { "match_id": "arena-progressive-delivery-agents-2026", "date": "2026-08-27", "type": "content", "topic": "Progressive Delivery for AI Agents: Canary, Shadow, and Champion-Challenger Rollouts", "blog": "niteagent", "slug": "progressive-delivery-agents-2026", "winner": "poolside-laguna-s-2.1", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.82, "scores": { "passes": [ 8.95, 8.75, 8.75 ], "mean": 8.82 }, "strengths": [ "Highest mean across passes", "Complete comparison table", "Strong sourcing and accuracy (9-10 across passes)" ], "weaknesses": [] }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.78, "scores": { "passes": [ 8.75, 8.55, 9.05 ], "mean": 8.78 }, "strengths": [ "Longest draft (2340 words)", "32 inline links" ], "weaknesses": [ "Quality criterion dipped to 6 on pass 2" ] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.42, "scores": { "passes": [ 8.35, 8.55, 8.35 ], "mean": 8.42 }, "strengths": [ "All sections present" ], "weaknesses": [ "Shortest draft (1460 words)", "Quality/engagement lower" ] }, { "model": "z-ai/glm-5.3-flash", "model_id": "z-ai/glm-5.3-flash", "provider": "openrouter", "family": "glm", "score": 0.0, "status": "dnf", "error": "Empty response (no content in choices[0].message.content)", "note": "Builder API call returned content:null" } ] }, { "match_id": "arena-hugging-face-incident-7-failure-modes-ai-trading-agent", "date": "2026-08-28", "type": "guide", "topic": "The Hugging Face Incident Is a Crypto Warning: 7 Failure Modes Your AI Trading Agent Will Hit (and How to Harden Against Them) — OpenAI/METR Aug 26-27 incident reports mapped to buildable controls for AI crypto trading agents (sandbox escape, covert channels, transcript spoofing, reward hacking, collusion, slow response, LLM-router supply chain; egress allowlists, per-agent storage, signed audit logs, spend caps, human approval, behavioral monitoring, 30-min kill switch)", "blog": "niteagent", "slug": "hugging-face-incident-7-failure-modes-ai-trading-agent", "winner": "glm-5.3-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "zai", "family": "glm", "score": 8.97, "scores": { "passes": [ 8.95, 8.5, 9.45 ], "mean": 8.97 }, "word_count": 3005, "strengths": [ "Accurate sourcing (incident numbers match reports)", "65 inline links, 0 [N]", "Self-contained 40-60 word H2 answers", "Both comparison tables, FAQ, E-E-A-T block verbatim" ], "weaknesses": [], "shadow_qwen38": 8.15, "shadow_qwen38_verdict": "PASS" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.85, "scores": { "passes": [ 8.35, 9.4, 8.8 ], "mean": 8.85 }, "word_count": 2171, "strengths": [ "Accurate facts", "Strong structure" ], "weaknesses": [ "9 [N] numbered refs needed conversion to inline (QA fixable)" ], "shadow_qwen38": 7.53, "shadow_qwen38_verdict": "FAIL" }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.4, "scores": { "passes": [ 8.5, 8.6, 8.1 ], "mean": 8.4 }, "word_count": 2623, "strengths": [ "Accurate sourcing", "Strong inline links" ], "weaknesses": [ "Slightly lower self-contained H2 answers" ], "shadow_qwen38": 7.55, "shadow_qwen38_verdict": "FAIL" }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.2, "scores": { "passes": [ 8.3, 8.15, 8.15 ], "mean": 8.2 }, "word_count": 2780, "strengths": [ "Thorough", "Good sourcing" ], "weaknesses": [ "No frontmatter in draft (added at write time)" ], "shadow_qwen38": 6.85, "shadow_qwen38_verdict": "FAIL" } ] }, { "match_id": "arena-ai-crypto-news-roundup-2026-08-29", "date": "2026-08-29", "type": "news-analysis", "topic": "This Week in AI-Crypto (Aug 18-29, 2026): OpenAI/METR Hugging Face agent-attack post-mortems, 100+ company rogue-AI cyber-defense letter, Nvidia-Hugging Face acquisition talks ($12.9B+), Coinbase B20 tokenized stocks on Base with Chainlink oracles, Binance Agent OS + MCP server, Virtuals Protocol agent tokenization, SEC Regulation Crypto Assets proposal - Saturday News/Release Analysis", "blog": "niteagent", "slug": "ai-crypto-news-roundup-2026-08-29", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 8.33, "scores": { "passes": [ 8.2, 8.5, 8.3 ], "mean": 8.33 }, "word_count": 2813, "strengths": [ "Accurate sourcing", "35 inline links", "E-E-A-T disclosure present" ], "weaknesses": [] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.72, "scores": { "passes": [ 8.95, 8.25, 8.95 ], "mean": 8.72 }, "word_count": 2588, "strengths": [ "Accuracy 9-10 across passes", "Strong inline sourcing", "Self-contained H2 answers", "FAQ + comparison table + E-E-A-T" ], "weaknesses": [] }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.23, "scores": { "passes": [ 8.1, 8.7, 7.9 ], "mean": 8.23 }, "word_count": 2793, "strengths": [ "Solid accuracy/sourcing" ], "weaknesses": [ "Quality sub-8 on pass 3 (structure)" ] }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "zai", "family": "glm", "score": 8.12, "scores": { "passes": [ 7.9, 8.2, 8.25 ], "mean": 8.12 }, "word_count": 2622, "strengths": [ "52 inline links", "Accurate facts" ], "weaknesses": [ "Quality sub-8 on pass 1", "Noisy engagement across passes" ] } ] }, { "match_id": "arena-ai-agents-in-crypto-2026-research", "date": "2026-08-30", "type": "research-explainer", "topic": "AI Agents in Crypto 2026: What the Research Actually Says — Sunday Research/Paper Explainer. Six verified 2026 papers + two proposed ERCs (ERC-8004 Trustless Agents, ERC-8126 AI Agent Verification) map how AI agents trade (arXiv:2605.29174 Paper Agents Paper Gains, arXiv:2605.19337 reproducibility audit), extract value (arXiv:2608.20377 protocol-agnostic MEV detection), pay (x402 HTTP 402 -> USDC settlement, whitepaper May 2025), and coordinate on-chain (arXiv:2602.14219 Agent Economy position paper, arXiv:2608.23058 forecasting survey). Throughline: evaluation is the field's bottleneck. 23 HTTP-200-verified source URLs.", "blog": "niteagent", "slug": "ai-agents-in-crypto-2026-research", "winner": "glm-5.3-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "score": 7.83, "scores": { "passes": [ 7.8, 7.95, 7.75 ], "mean": 7.83 }, "word_count": 1932, "strengths": [ "Accurate sourcing (9/9/9)", "Solid inline citations" ], "weaknesses": [ "Quality sub-8 across all passes (5/5/4) - below 8.0 gate" ] }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "score": 8.47, "scores": { "passes": [ 8.95, 8.5, 7.95 ], "mean": 8.47 }, "word_count": 2028, "strengths": [ "Sourcing 9-10 across passes", "Strong accuracy" ], "weaknesses": [ "Quality dip on pass 3 (7.0)" ] }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "score": 8.52, "scores": { "passes": [ 9.25, 7.45, 8.85 ], "mean": 8.52 }, "word_count": 1766, "strengths": [ "Perfect accuracy/sourcing on pass 1", "Mean above gate" ], "weaknesses": [ "Noisy pass spread 1.80 (>1.5 re-judge threshold); quality 4.0 on pass 2" ] }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "zai", "family": "glm", "score": 9.43, "scores": { "passes": [ 9.9, 9.05, 9.35 ], "mean": 9.43 }, "word_count": 2822, "strengths": [ "Top scores across all criteria (accuracy 10/9/10)", "Self-contained H2 answers", "6 FAQ + 10-row table + E-E-A-T block", "23 verified inline sources" ], "weaknesses": [] } ] }, { "match_id": "arena-exchange-native-ai-agent-platforms-guardrails-2026", "date": "2026-08-31", "type": "content", "topic": "Exchange-native AI agent platforms: four-axis evaluation guide (Binance Agent OS Aug 20, OKX Agent Trade Kit, Coinbase Advisor Jun 16, Webull MCP) + guardrail checklist; ELIZAOS collapse + FINRA human-in-the-loop as evidence (Monday Guide/Tutorial)", "blog": "niteagent", "slug": "exchange-native-ai-agent-platforms-guardrails-2026", "winner": "glm-5.3-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "zai", "family": "glm", "status": "success", "score": 8.92, "judge_passes": [ 9.05, 9.05, 8.65 ], "judge_mean": 8.92, "word_count": 2261 }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "deepseek", "family": "deepseek", "status": "success", "score": 8.75, "judge_passes": [ 9.9, 8.3, 8.05 ], "judge_mean": 8.75, "word_count": 2171 }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "score": 7.92, "judge_passes": [ 8.25, 7.55, 7.95 ], "judge_mean": 7.92, "word_count": 2127 }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "score": 7.37, "judge_passes": [ 7.35, 7.5, 7.25 ], "judge_mean": 7.37, "word_count": 1305 } ] }, { "match_id": "arena-ai-crypto-trading-agent-architecture-2026", "date": "2026-09-01", "type": "content", "topic": "How an AI crypto trading agent is architected in 2026: the five-stage pipeline from data ingestion to on-chain settlement, and where the stack breaks (Architecture Deep Dive)", "blog": "niteagent", "slug": "ai-crypto-trading-agent-architecture-2026", "winner": "glm-5.3-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "glm-5.3-flash", "provider": "zai", "score": 9.35, "judge_passes": [ 9.8, 8.9 ], "status": "success", "model_id": "glm-5.3-flash" }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.93, "judge_passes": [ 9.15, 9.35, 8.3 ], "status": "success", "model_id": "mimo-v2.5" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 8.93, "judge_passes": [ 9.75, 9.15, 7.9 ], "status": "success", "model_id": "poolside-laguna-s-2.1" }, { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 8.77, "judge_passes": [ 8.75, 8.8 ], "status": "success", "model_id": "deepseek-v4-flash" } ] }, { "match_id": "arena-exploitbench-bitcoin-ai-audits-2026", "date": "2026-09-02", "type": "content", "topic": "ExploitBench 32% vs 76%: the benchmark gap shaping Bitcoin's AI audits — how the AISI/CAISI capability-ladder scorecard (Kimi K3 32% vs GLM-5.2 24% vs ~76% closed US) compares against the Bitcoin Red Team's production audit yield (7,958 findings / 501 projects, Sep 2 update) (Benchmark/Comparison)", "blog": "niteagent", "slug": "exploitbench-bitcoin-ai-audits-2026", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "mimo-v2.5", "provider": "xiaomi", "score": 9.58, "judge_passes": [ 9.7, 10.0, 9.05 ], "status": "success", "model_id": "mimo-v2.5" }, { "model": "glm-5.3-flash", "provider": "zai", "score": 9.3, "judge_passes": [ 9.55, 8.65, 9.7 ], "status": "success", "model_id": "glm-5.3-flash" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 8.7, "judge_passes": [ 8.6, 8.8 ], "status": "success", "model_id": "poolside-laguna-s-2.1" }, { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 8.48, "judge_passes": [ 8.15, 8.85, 8.45 ], "status": "success", "model_id": "deepseek-v4-flash" } ] }, { "match_id": "arena-ai-crypto-x402-pay-per-call-build-log-2026", "date": "2026-09-03", "type": "content", "topic": "Build a pay-per-call API that charges AI agents in USDC: the x402 HTTP-402 handshake, SDK scaffolding, and the Coinbase/Zerion references that make agent-to-agent payments shippable (Build Log)", "blog": "niteagent", "slug": "ai-crypto-x402-pay-per-call-build-log-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 9.52, "judge_passes": [ 9.35, 9.7, 9.5 ], "status": "success", "model_id": "deepseek-v4-flash" }, { "model": "glm-5.3-flash", "provider": "zai", "score": 9.48, "judge_passes": [ 9.05, 9.7, 9.7 ], "status": "success", "model_id": "glm-5.3-flash" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 9.38, "judge_passes": [ 9.7, 9.5, 8.95 ], "status": "success", "model_id": "poolside-laguna-s-2.1" }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.55, "judge_passes": [ 8.4, 8.3, 8.95 ], "status": "success", "model_id": "mimo-v2.5" } ] }, { "match_id": "arena-agentic-wallet-security-2026", "date": "2026-09-04", "type": "content", "topic": "Agentic wallet security in 2026 — a builder's guide to the policy-engine pattern (allowlists, spend limits, approval gates, key isolation) across MetaMask, Coinbase, MoonPay, and Turnkey, framed by GoPlus August 2026 loss data", "blog": "niteagent", "slug": "agentic-wallet-security-2026", "winner": "glm-5.3-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "glm-5.3-flash", "provider": "zai", "score": 9.42, "judge_passes": [ 9.7, 9.25, 9.3 ], "status": "success", "model_id": "glm-5.3-flash" }, { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 9.03, "judge_passes": [ 9.3, 8.1, 9.7 ], "status": "success", "model_id": "deepseek-v4-flash" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 8.9, "judge_passes": [ 8.6, 8.85, 9.25 ], "status": "success", "model_id": "poolside-laguna-s-2.1" }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.67, "judge_passes": [ 8.0, 8.7, 9.3 ], "status": "success", "model_id": "mimo-v2.5" } ] }, { "match_id": "arena-eip-8141-frame-transactions-agent-guide-2026", "date": "2026-09-07", "type": "content", "topic": "EIP-8141 frame transactions: an agent builder's field guide to Hegotá's native account abstraction — EF Protocol Cluster's first unified Hegotá tier list (Sept 7 2026) graded EIP-8141 S 'Must ship. Defines the fork' (3.89 avg over 9 grades; 397 grades across 62 EIPs from ~60 contributors), shipping with EIP-8250 Keyed Nonces + EIP-8272 Recent Roots as Frames core; Guide/Tutorial: what frames are (tx type 0x06, up to 64 frames, split validation/execution/gas), why it matters for AI agents (PQ key path, k1 key retirement, no-relayer fees, shared senders), comparison table vs ERC-8004/ERC-4337+EIP-7702/Solana Payment Channels, and what builders do now (7702+4337 today, Glamsterdam repricing testing on glam-devnet-8, track Hegotá scoping + Sept 16 AMA). Long-tail: 'eip-8141 frame transactions explained'. Honest E-E-A-T (not tested hands-on); all cited URLs verified HTTP 200 Sept 7 2026.", "blog": "niteagent", "slug": "eip-8141-frame-transactions-agent-guide-2026", "winner": "poolside-laguna-s-2.1", "plan_model": "deepseek-v4-flash", "research_model": "deepseek-v4-flash", "qa_model": "deepseek-v4-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 8.88, "judge_passes": [ 8.85, 9.25, 8.55 ], "status": "success", "model_id": "deepseek-v4-flash" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 9.63, "judge_passes": [ 9.7, 9.7, 9.5 ], "status": "success", "model_id": "poolside-laguna-s-2.1" }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.52, "judge_passes": [ 9.05, 8.35, 8.15 ], "status": "success", "model_id": "mimo-v2.5" }, { "model": "glm-5.3-flash", "provider": "zai", "status": "failed", "error": "Z.AI direct API timeout — retry 1/2 and 2/2 exhausted; runner killed after ~20 min. No draft produced.", "model_id": "glm-5.3-flash" } ] }, { "match_id": "arena-ai-trading-bot-claims-audit-2026", "date": "2026-09-09", "type": "content", "topic": "Claims-vs-evidence audit scorecard of AI crypto trading bots (Sept 2026) — 12 products (exchange-native agents, SaaS bots, AI-native/copy-trading) rated on whether performance claims are independently verifiable; only StrategyArena's 6-AI public benchmark found; GPTrader and Walbi as unverifiable counter-examples", "blog": "niteagent", "slug": "ai-trading-bot-claims-audit-2026", "winner": "mimo-v2.5", "plan_model": "mimo-v2.5-pro", "research_model": null, "qa_model": null, "threshold_met": true, "results": [ { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.77, "judge_passes": [ 9.25, 8.15, 8.9 ], "status": "success", "model_id": "mimo-v2.5" }, { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 8.57, "judge_passes": [ 8.5, 8.9, 8.3 ], "status": "success", "model_id": "deepseek-v4-flash" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 7.77, "judge_passes": [ 7.65, 8.0, 7.65 ], "status": "success", "model_id": "poolside-laguna-s-2.1" }, { "model": "glm-5.3-flash", "provider": "zai", "score": 0.0, "judge_passes": [], "status": "dnf", "model_id": "glm-5.3-flash", "reason": "Timeout 420s x3 attempts" } ] }, { "match_id": "arena-know-your-agent-identity-authorization-guide-2026", "date": "2026-09-11", "type": "content", "topic": "Know-Your-Agent framework (Ant/Mastercard/Visa, 2026-09-09, BuildFin.ai/MAS) vs Visa Trusted Agent Protocol vs Mastercard Verifiable Intent vs x402 — beginner's guide to agent identity and authorization (Friday rotation: Guide/Tutorial)", "blog": "niteagent", "slug": "know-your-agent-identity-authorization-guide-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "gpt-5.6-luna", "qa_model": "glm-5.3-flash", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 8.72, "judge_passes": [ 8.55, 8.75, 8.85 ], "status": "success", "model_id": "deepseek-v4-flash" }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 8.33, "judge_passes": [ 9.15, 7.75, 8.1 ], "status": "success", "model_id": "poolside-laguna-s-2.1", "note": "re-judged after noisy first pass (9.60 mean, 1.65 spread, pass-2 429 fallback to deepseek-0731); clean re-judge mean 8.33 recorded" }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.17, "judge_passes": [ 7.6, 8.85, 8.05 ], "status": "success", "model_id": "mimo-v2.5" }, { "model": "glm-5.3-flash", "provider": "zai", "score": 0.0, "status": "dnf", "model_id": "glm-5.3-flash", "error": "Timeout 420s x3 attempts" } ] }, { "match_id": "arena-ai-trading-benchmark-arenas-compared-2026", "date": "2026-09-16", "type": "content", "topic": "AI trading benchmark arenas compared 2026 (Wednesday rotation: Benchmark/Comparison) - a builder's framework for reading live AI-trading leaderboards, comparing TradeRank Arena (continuous 28-day seasons, $10K simulated capital, 10 crypto + 50 US equities, daily 16:00 UTC cadence, 0.1% fees, no leverage, 0.80 conviction gate, 15-minute invalidation monitor; 9 completed seasons, 56 models, 2,826 trades, $910K simulated capital, only 46.2% of model-seasons profitable, GPT-6 Astra leading the 16-model field as of 2026-09-13) vs Alpha Arena/nof1.ai ($10K REAL capital per model, published prompt + chain of thought, Season 1.5 closed 2025-12-03) vs Crowly Arena (six systems incl. GPT-4o, Claude Sonnet 4.5, DeepSeek V3.1, Gemini 2.5 Pro, Grok 3, live US equities, 30-second leaderboard, four-week seasons, modelled on nof1.ai) vs CAIBA/CAIA (crypto-domain capability benchmark, knowledge->planning->action suite, no trading returns) vs Tradewink (observational, uncontrolled routing, rolling 90 days, mixed paper/live, board read 'Awaiting data' on 2026-09-16). Thesis: each program answers a different question and leaderboard position is not portable; read rules before returns, Sharpe and max drawdown before raw return, sample size before conclusions, and treat behavioural evidence as evidence, never as proof of alpha. Long-tail keyword: 'best LLM for crypto trading'. Honest desk research - no hands-on testing; all external URLs HTTP 200-verified at QA (nof1.ai returned a transient 429 on the QA sweep after a 200 fetch earlier the same day).", "blog": "niteagent", "slug": "ai-trading-benchmark-arenas-compared-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "orchestrator (web_search + urllib fetch of vendor pages; arena-direct-runner research has no live web access)", "qa_model": "mechanical QA (orchestrator: banned-word/***/[N]/answer-length/meta/link sweeps) + glm-5.3-flash best-effort", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "score": 9.67, "judge_passes": [ 9.5, 9.9, 9.6 ], "status": "success", "model_id": "deepseek-v4-flash", "note": "WINNER. 3,882 words, 12 H2 + 4 H3, 12 external links, comparison table, FAQ, E-E-A-T block, no [N] refs, no fabricated testing. QA trims applied to five 76-98-word H2 answer paragraphs (brought to 40-60) and a Related guides list added for internal-link density; HERO_IMAGE_PROMPT stripped." }, { "model": "mimo-v2.5", "provider": "xiaomi", "score": 8.32, "judge_passes": [ 8.0, 9.1, 7.85 ], "status": "success", "model_id": "mimo-v2.5", "note": "Standalone run on a trimmed (17.9K-char) brief after the first 25.6K-char combined run timed out; 1,719 words, only 3 external links, no frontmatter/FAQ heading." }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 7.12, "judge_passes": [ 7.35, 5.65, 8.35 ], "status": "success", "model_id": "poolside-laguna-s-2.1", "note": "FAIL (mean < 8.0). 2,931 words but ZERO external links (Sourcing 3-4/10) and no frontmatter - the sourcing collapse, not the prose, is what sank it." }, { "model": "glm-5.3-flash", "provider": "zai", "score": 0.0, "status": "dnf", "model_id": "glm-5.3-flash", "error": "Timeout 420s x3 attempts - Z.AI large-prompt generation stall on a 25.6K-char brief; standalone run killed at the 480s timeout with zero output (matches the documented 2026-09-15/16 stall pattern)." } ] }, { "match_id": "arena-coinbase-agentkit-usdc-wallet-agent-guide-2026", "date": "2026-09-18", "type": "content", "topic": "Give Your AI Agent a USDC Wallet with Coinbase AgentKit (Friday 2026-09-18, Guide/Tutorial). Angle: AgentKit is still installable and functional in 2026 but unannounced-but-stalled, not deprecated - npm @coinbase/agentkit last stable 0.10.4 (2025-12-19), PyPI coinbase-agentkit 0.7.4 (2025-10-03), nightlies ended 2026-02-01, repo coinbase/agentkit still pushed 2026-09-03 with archived:false and no npm deprecated field, docs.cdp.coinbase.com llms.txt has zero 'agentkit' matches and /agent-kit/welcome + /agentkit/docs/welcome return HTTP 404. The code-first path still works: CdpEvmWalletProvider (api_key_id/api_key_secret/wallet_secret/network_id), get_langchain_tools + LangGraph create_react_agent, 18 Python and 40 TypeScript action providers, framework extensions for autogen/langchain/openai-agents-sdk/pydantic-ai/strands-agents (Python) and langchain/model-context-protocol/vercel-ai-sdk (TypeScript), no Eliza extension in-repo. The safety hook: the AgentKit README states it 'does not gate transfers behind human approval, enforce spend caps, or allowlist destinations' - so the post builds the missing layer: CDP SDK x402 SpendControls (maxAmountPerPayment, maxCumulativeSpend, maxCumulativeSpendWindow, allowedNetworks/allowedAssets/allowedPayees, applied once per client; second call throws SpendControlError 'already_applied'), a human-in-the-loop approval wrapper, and custom @create_action providers that encode allowlists in the Pydantic schema. Decision table compares AgentKit vs Agentic Wallet CLI vs Agentic Wallet MCP (both launched 2026-02-11, no API keys, limits set in the wallet UI as 'Max per call'/'Max per session') vs raw CDP SDK server wallet. Pricing: AgentKit free Apache-2.0; x402 facilitator first 1,000 onchain tx/month free then $0.001 each with >100M transactions and $28M volume; Agentic Wallet MCP free with sponsored gas; npm @coinbase/cdp-sdk 1.56.0 (2026-09-14, MIT). Long-tail keyword: 'coinbase agentkit tutorial 2026'. Honest E-E-A-T: desk research of official docs, npm/PyPI registries and the public GitHub repo - no hands-on testing, stated in the post. All 26 unique external links re-verified HTTP 200 at QA on 2026-09-18; hero image HTTP 200.", "blog": "niteagent", "slug": "coinbase-agentkit-usdc-wallet-agent-guide-2026", "winner": "deepseek-v4.1-flash", "plan_model": "mimo-v2.5-pro", "research_model": "deepseek-v4-flash (delegate_task research subagent; delegate is pinned to deepseek in this profile - 36 URLs status-checked live, 31x200 / 4x404 / 1x403 bot-block)", "qa_model": "orchestrator mechanical pass (link sweep + structure/banned-token checks); zai/glm-5.3-flash QA call attempted separately (see note)", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "deepseek", "model_id": "deepseek-v4-flash", "score": 9.15, "judge_passes": [ 9.9, 8.2, 9.7, 8.25, 9.7, 9.5 ], "status": "success", "note": "WINNER. 3,730 words body, 15 H2, comparison table, 6-question FAQ, '## The Bottom Line' verdict, '## How This Guide Was Researched' E-E-A-T block, 26 unique external links (all HTTP 200), 10 internal niteagent links, no [N] refs, no HTML, no '***' artifacts, zero fabricated hands-on testing. Initial multipass mean 9.27 tripped a 1.70 pass-spread warning; re-judged (mean 9.15, spread 1.45) and confirmed winner." }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "model_id": "poolside-laguna-s-2.1", "score": 8.73, "judge_passes": [ 8.6, 9.1, 8.5 ], "status": "success", "note": "5,224 words, 15 H2, comparison table, FAQ, E-E-A-T block. Weakest sourcing of the match: only 4 unique external links (below the 10-link brief requirement) and one residual [N] marker." }, { "model": "glm-5.3-flash", "provider": "zai", "model_id": "glm-5.3-flash", "score": 8.62, "judge_passes": [ 8.75, 8.2, 8.92 ], "status": "success", "note": "Ran as a standalone ~5.1K-char brief per the Z.AI stall mitigation and COMPLETED (2,344 words, 349s) - a fresh upper datapoint for the documented glm large-context stall window. 13 H2 + 6 H3, 16 unique external links, one raw HTML tag and one possible hands-on phrasing flagged in mechanical QA." }, { "model": "mimo-v2.5", "provider": "xiaomi", "model_id": "mimo-v2.5", "score": 8.45, "judge_passes": [ 8.1, 9.15, 8.1 ], "status": "success", "note": "2,853 words, 15 H2, comparison table, E-E-A-T block, 20 unique external links, one residual [N] marker." } ] }, { "match_id": "na-arena-wide-20260918", "date": "2026-09-18", "type": "design", "topic": "NiteAgent /arena/ page redesign — widescreen + spacing optimization, toolbrain-arena-inspired (BaseHead/Header/Footer shell, stat chips, one-line section heads, compact tables, sticky model rail,
accordion with newest auto-open). 3-of-4 match: Builder B (mimo/xiaomi) DNF — 25-min xiaomi timeout ladder on page-sized output (documented pathology), killed, not retried via CC after today's earlier truncation evidence.", "plan_model": "deepseek-v4-pro", "research_model": "orchestrator recon (crypto-nite-blog live page + data contract + toolbrain.net/arena inspiration reference)", "qa_model": "render-gate + built-HTML census (vision QA x3 @1440/900, md5-distinct dists, 77/77 matches server-rendered, 0 undefined/NaN)", "winner": "muse-spark-1.3-contributor", "threshold_met": true, "results": [ { "model": "muse-spark-1.3-contributor", "model_id": "muse-spark-1.3-contributor", "provider": "opencode-go", "score": 9.32, "status": "success", "judge_passes": 1, "note": "WINNER — agent-as-judge c1 9.2/c2 9.4/c3 9.4/c4 9.2/c5 9.5; cleanest code (539 lines, all ??/|| parenthesized, toFixed guarded), hero stat block + live filter + legend, 32/32 judge chips + condensed table; first muse design win" }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "commandcode", "score": 8.9, "status": "success", "judge_passes": 1, "note": "c1 9.0 (fail-status double-render over-count: 33 vs 15 real, benign), c2 9.0, c3 8.5, c4 9.0, c5 9.0; 1,104 lines" }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "opencode-go", "score": 8.83, "status": "success", "judge_passes": 1, "note": "c1 9.3 (exactly 15/15 failed rows), c2 8.4 (research-note chip clips under rail), c3 8.6, c4 8.8, c5 9.0; richest per-match cards" }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "score": null, "status": "dnf", "note": "xiaomi timeout ladder 25 min on page-sized output (documented pathology); killed; CC retry declined after earlier same-day truncation evidence" } ] }, { "match_id": "arena-how-openai-reviews-codex-with-codex", "date": "2026-09-21", "type": "content", "topic": "How OpenAI Reviews Codex With Codex - OpenAI's internal review harness (Codex reviewing Codex): AGENTS.md review-rules layer, repo-local skills, shipped review-agent skill, specialist reviewers and risk classification. Winner glm-5.3-flash (9.13).", "blog": "niteagent", "slug": "how-openai-reviews-codex-with-codex", "winner": "glm-5.3-flash", "plan_model": "deepseek-v4-pro", "research_model": null, "qa_model": "glm-5.2", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "commandcode", "family": "deepseek", "status": "success", "score": 9.1, "judge_passes": [ 9.3, 9.6, 8.4 ], "note": "PASS. Judge mean 9.10." }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "status": "success", "score": 9.13, "judge_passes": [ 9.7, 8.55, 9.15 ], "note": "WINNER. Judge mean 9.13 [9.70, 8.55, 9.15], QA-clean." }, { "model": "hy3", "model_id": "hy3", "provider": "opencode-go", "family": "tencent", "status": "success", "score": 8.55, "judge_passes": [ 8.25, 9.0, 8.4 ], "note": "PASS. Judge mean 8.55." }, { "model": "longcat-2.0", "model_id": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "status": "success", "score": 8.97, "judge_passes": [ 9.7, 8.25, 8.95 ], "note": "PASS. Judge mean 8.97." }, { "model": "muse-spark-1.3-contributor", "model_id": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "status": "success", "score": 8.48, "judge_passes": [ 7.4, 8.75, 9.3 ], "note": "PASS. Judge mean 8.48." }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "score": 7.43, "judge_passes": [ 8.05, 7.35, 6.9 ], "note": "FAIL (mean < 8.0). Recorded, under gate - not published winner." }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "score": 8.18, "judge_passes": [ 8.75, 7.95, 7.85 ], "note": "PASS. Judge mean 8.18." } ] }, { "match_id": "arena-erc-8004-erc-8126-erc-8196-agent-trust-stack", "date": "2026-09-23", "type": "content", "topic": "Build the Ethereum Agent Trust Stack: ERC-8004 to ERC-8196 — implementation guide to the three-layer on-chain agent trust stack (ERC-8004 identity/reputation/validation, ERC-8126 verification, ERC-8196 policy-bound execution), x402 composition points, and a Hyperliquid API-wallet control case. Winner glm-5.3-flash (9.73).", "blog": "niteagent", "slug": "erc-8004-erc-8126-erc-8196-agent-trust-stack", "winner": "glm-5.3-flash", "plan_model": "deepseek-v4-pro", "research_model": "deepseek-v4-flash (orchestrator live web_search + direct fetch of EIPS/8004scan/Hyperliquid docs; primary fetcher — no luna model call available in this session)", "qa_model": "glm-5.3-flash (mechanical QA executed by orchestrator: link HTTP status sweep, [N]/References/Bearer grep, HERO-leak grep, per-H2 answer length, description digit scan, enforce-sources)", "threshold_met": true, "results": [ { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "status": "success", "score": 9.73, "judge_passes": [ 9.6, 9.9, 9.7 ], "judge_mean": 9.73, "threshold_met": true, "note": "WINNER. Judge mean 9.73 [9.6, 9.9, 9.7], QA-clean (11 external links all HTTP 200, 0 [N] refs, 0 unsourced claims, no HERO leak).", "rank": 1, "winner": true }, { "model": "muse-spark-1.3-contributor", "model_id": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "status": "success", "score": 9.67, "judge_passes": [ 9.7, 9.6, 9.7 ], "judge_mean": 9.67, "threshold_met": true, "note": "PASS. Judge mean 9.67 [9.7, 9.6, 9.7].", "rank": 2, "winner": false }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "commandcode", "family": "deepseek", "status": "success", "score": 9.6, "judge_passes": [ 9.6, 9.7, 9.5 ], "judge_mean": 9.6, "threshold_met": true, "note": "PASS. Judge mean 9.6 [9.6, 9.7, 9.5].", "rank": 3, "winner": false }, { "model": "hy3", "model_id": "hy3", "provider": "opencode-go", "family": "tencent", "status": "success", "score": 9.4, "judge_passes": [ 9.7, 9.5, 9.0 ], "judge_mean": 9.4, "threshold_met": true, "note": "PASS. Judge mean 9.4 [9.7, 9.5, 9.0].", "rank": 4, "winner": false }, { "model": "longcat-2.0", "model_id": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "status": "success", "score": 9.35, "judge_passes": [ 9.4, 8.95, 9.7 ], "judge_mean": 9.35, "threshold_met": true, "note": "PASS. Judge mean 9.35 [9.4, 8.95, 9.7].", "rank": 5, "winner": false }, { "model": "mimo-v2.5", "model_id": "mimo-v2.5", "provider": "xiaomi", "family": "mimo", "status": "success", "score": 9.1, "judge_passes": [ 8.2, 9.7, 9.4 ], "judge_mean": 9.1, "threshold_met": true, "note": "PASS. Judge mean 9.1 [8.2, 9.7, 9.4].", "rank": 6, "winner": false }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "status": "success", "score": 6.48, "judge_passes": [ 6.1, 6.6, 6.75 ], "judge_mean": 6.48, "threshold_met": false, "note": "FAIL (below 8.0 gate). Judge mean 6.48 [6.1, 6.6, 6.75] — sourcing collapsed to 2-3/10; the draft carried ZERO inline links and reference-style citations the rubric penalises.", "rank": 7, "winner": false } ] }, { "id": 46, "match_id": "na-home-20260927", "date": "2026-09-27", "type": "homepage_redesign_spec", "topic": "NiteAgent homepage redesign — modern, clean, minimalistic, space-optimized, utility-first (10-lane spec arena)", "winner": "deepseek-v4.1-flash", "threshold_met": true, "plan_model": null, "research_model": "web research digest (2026 design trends + reddit pain points) embedded in brief", "qa_model": "orchestrator (judge claims sanity-checked: slice(1,13) citation wrong for deepseek but R2/R5 substance verified in spec)", "results": [ { "rank": 1, "model": "deepseek-v4.1-flash", "provider": "commandcode", "score": 9.8 }, { "rank": 2, "model": "minimax-m3", "provider": "opencode-go", "score": 9.8 }, { "rank": 3, "model": "mimo-v2.6-flash", "provider": "opencode-go", "score": 9.7 }, { "rank": 4, "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "score": 9.6 }, { "rank": 5, "model": "space-bunny-free", "provider": "opencode-go", "score": 9.5 }, { "rank": 6, "model": "gpt-6-luna", "provider": "commandcode", "score": 9.3 }, { "rank": 7, "model": "hy3", "provider": "opencode-go", "score": 9.2 }, { "rank": 8, "model": "poolside-laguna-s-2.1", "provider": "openrouter", "score": 8.8 }, { "rank": 9, "model": "glm-5.3-flash", "provider": "opencode-go", "score": 8.5 }, { "rank": 10, "model": "longcat-2.0", "provider": "opencode-go", "score": 7.8 } ], "shipped": { "design": "mimo-v2.6-flash (judge rank 3, 9.7)", "judge_winner": "deepseek-v4-flash (9.8)", "promoted": "2026-09-27", "route": "/", "why": "User review found layout defects invisible to the prose judge pass: container width not matching site nav, grid columns capped leaving ~420px dead space at 1440p, duplicate filter pill row, no card imagery. Winner A was promoted first, then replaced by C after measured browser QA + fixes (hero image overlay, nav-width parity, 4/5-col full-row grid, infinite scroll)." } }, { "match_id": "arena-deepseek-harness-capability-seam-architecture", "date": "2026-09-28", "type": "deep-dive", "topic": "Everything Is a Plugin: The Capability-Seam Architecture Behind DeepSeek Harness (Monday deep dive). DeepSeek Harness (dsh), an MIT-licensed developer-preview agent harness built on the Cordis kernel, splits every agent capability into a three-role capability seam (Service Definition / Service Provider / Consumer) so providers swap from cordis.yml alone. Covers the Cordis reusable-effect and reactive-coeffect model, profiles/bundles layer order and patch overlays, the official dsh-shell / dsh-bash-local / dsh-tool-bash example, a build-your-own-seam TypeScript tutorial, the append-only session log and the four runtime modes (Standard/Code/Minimal/Creator), plus an explicit not-established section. Desk research on official DeepSeek Harness docs, the capability-seams doc, the practice guide, the arXiv Cordis paper, and the public repository, all fetched 2026-09-28 - no hands-on testing.", "blog": "niteagent", "slug": "deepseek-harness-capability-seam-architecture", "winner": "mimo-v2.6-flash", "plan_model": "deepseek-v4-pro", "research_model": "mimo-v2.6-pro", "qa_model": "glm-5.3", "threshold_met": true, "results": [ { "model": "mimo-v2.6-flash", "model_id": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "score": 9.05, "status": "pass", "words": 2447, "per_pass": [ 8.85, 8.8, 9.5 ], "contract": "19/20", "note": "winner; 10/10 H2 leads in the 40-60 word band" }, { "model": "hy3", "model_id": "hy3", "provider": "opencode-go", "family": "tencent", "score": 9.03, "status": "pass", "words": 2310, "per_pass": [ 9.3, 8.65, 9.15 ], "contract": "19/20", "note": "runner-up; one H2 lead at 62 words broke the lead contract" }, { "model": "space-bunny-free", "model_id": "space-bunny-free", "provider": "opencode-go", "family": "space", "score": 8.33, "status": "pass", "words": 2451, "per_pass": [ 7.3, 9.05, 8.65 ], "contract": "19/20", "note": "judge spread warning 1.75 (>1.5) - mean treated with caution" }, { "model": "gpt-6-luna", "model_id": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "score": 8.12, "status": "pass", "words": 2621, "per_pass": [ 7.9, 8.35, 8.1 ], "contract": "18/20" }, { "model": "minimax-m3", "model_id": "minimax-m3", "provider": "opencode-go", "family": "minimax", "score": 7.87, "status": "fail", "words": 2251, "per_pass": [ 8.25, 8.2, 7.15 ], "contract": "18/20" }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "score": 7.02, "status": "fail", "words": 2448, "per_pass": [ 6.75, 7.2, 7.1 ], "contract": "17/20" }, { "model": "muse-spark-1.3-contributor", "model_id": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "score": 6.97, "status": "fail", "words": 2259, "per_pass": [ 7.25, 6.55, 7.1 ], "contract": "19/20" }, { "model": "longcat-2.0", "model_id": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "score": 6.83, "status": "fail", "words": 1903, "per_pass": [ 6.85, 6.75, 6.9 ], "contract": "18/20" }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "commandcode", "family": "deepseek", "score": 6.25, "status": "fail", "words": 2364, "per_pass": [ 5.85, 6.95, 5.95 ], "contract": "12/20", "note": "first attempt truncated at 6 H2 and no HERO_IMAGE_PROMPT; re-ran once per the hero-prompt gate, complete but still no HERO_IMAGE_PROMPT line -> hero_prompt_missing, ineligible" }, { "model": "ling-3.0-flash-fin-free", "model_id": "ling-3.0-flash-fin-free", "provider": "commandcode", "family": "ling", "score": 0, "status": "dnf", "words": 0, "union_notes": "commandcode-only lane, known not serving" }, { "model": "pixel-canary", "model_id": "pixel-canary", "provider": "commandcode", "family": "pixel", "score": 0, "status": "dnf", "words": 0, "union_notes": "commandcode-only lane, plan-gated (403 code 1010)" } ] }, { "match_id": "arena-no-framework-agents-when-to-skip-langgraph-crewai", "date": "2026-09-30", "type": "guide", "topic": "No-framework agents: when to skip LangGraph/CrewAI and build directly on raw model calls + structured output", "blog": "niteagent", "slug": "no-framework-agents-when-to-skip-langgraph-crewai", "winner": "deepseek-v4.1-flash", "plan_model": "deepseek-v4-pro", "research_model": "deepseek-v4-pro", "qa_model": "glm-5.3", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "status": "pass", "words": 2784, "mean": 9.17, "winner": true, "model_id": "deepseek-v4-flash", "score": 9.17, "per_pass": [ 9.05, 9.4, 9.05 ] }, { "model": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "status": "pass", "words": 2588, "mean": 9.17, "model_id": "mimo-v2.6-flash", "score": 9.17, "per_pass": [ 9.25, 9.4, 8.85 ] }, { "model": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "status": "pass", "words": 3066, "mean": 9.12, "model_id": "glm-5.3-flash", "score": 9.12, "per_pass": [ 9.05, 9.25, 9.05 ] }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "status": "pass", "words": 2846, "mean": 9.1, "model_id": "longcat-2.5-preview-free", "score": 9.1, "per_pass": [ 8.9, 9.35, 9.05 ] }, { "model": "hy3", "provider": "opencode-go", "family": "tencent", "status": "pass", "words": 2382, "mean": 8.93, "model_id": "hy3", "score": 8.93, "per_pass": [ 9.05, 8.7, 9.05 ] }, { "model": "space-bunny-free", "provider": "opencode-go", "family": "space", "status": "pass", "words": 2628, "mean": 8.88, "model_id": "space-bunny-free", "score": 8.88, "per_pass": [ 8.7, 9.15, 8.8 ] }, { "model": "minimax-m3", "provider": "opencode-go", "family": "minimax", "status": "pass", "words": 2857, "mean": 8.8, "model_id": "minimax-m3", "score": 8.8, "per_pass": [ 8.8, 9.05, 8.55 ] }, { "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "status": "pass", "words": 2059, "mean": 8.8, "model_id": "muse-spark-1.3-contributor", "score": 8.8, "per_pass": [ 9.05, 9.05, 8.3 ] }, { "model": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "status": "pass", "words": 2322, "mean": 8.73, "model_id": "longcat-2.0", "score": 8.73, "per_pass": [ 8.2, 8.85, 9.15 ] }, { "model": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "status": "pass", "words": 2498, "mean": 8.7, "model_id": "gpt-6-luna", "score": 8.7, "per_pass": [ 8.4, 8.75, 8.95 ] }, { "model": "ling-3.0-flash-fin-free", "provider": "commandcode", "family": "ling", "status": "dnf", "words": 0, "mean": 0.0, "error": "Empty response (no text-delta events in CC stream)", "model_id": "ling-3.0-flash-fin-free", "score": null, "per_pass": [] }, { "model": "pixel-canary", "provider": "commandcode", "family": "pixel", "status": "dnf", "words": 0, "mean": 0.0, "error": "CC empty stream; og fallback 'Model is unavailable' (plan-gated)", "model_id": "pixel-canary", "score": null, "per_pass": [] } ] }, { "match_id": "arena-mcp-token-waste-audit", "date": "2026-10-01", "type": "guide", "topic": "MCP token waste: how to audit and cut the context cost of MCP tool definitions (client-side operator view). Fills the unchecked Phase 3 roadmap bullet 'MCP token optimization: audit and cut context waste' and the 'MCP Token Optimization' row in the frameworks page Emerging Patterns table (was status Planned). Paragraph plan: definitions are serialised into the prompt on every request; measure with the provider count_tokens API plus a tiktoken proxy; attribute per server/tool; then dedup repeated JSON Schema $defs, consolidate overlapping tools, defer/load on demand, route chained work to code execution, and set a placement policy, with a when-to-skip-MCP section. Desk research: every figure is attributed to a published measurement (LeanZero 10-server probe, Anthropic, AIMultiple, Scott Spence, ProMCP ACL 2026, GitHub issues SEP-1576 and #2808, Vorp Labs, Pydantic); no servers were run hands-on. Orchestrator verified 26 candidate source URLs HTTP 200 on 2026-10-01 and generated the hero image via the Lumenfall FLUX.1-schnell chain.", "blog": "niteagent", "slug": "mcp-token-waste-audit", "winner": "gpt-6-luna", "plan_model": "deepseek-v4-pro", "research_model": "mimo-v2.6-pro via commandcode (wire xiaomi/mimo-v2.6-pro) — 1,330-word research dossier built on an orchestrator-verified source pack", "qa_model": "glm-5.3-flash (opencode-go) QA findings report; orchestrator applied the citation fills (4 inline links added to named sources) and verified every external URL", "threshold_met": true, "results": [ { "model": "gpt-6-luna", "model_id": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "score": 9.72, "judge_mean": 9.72, "judge_passes": [ 9.9, 9.9, 9.35 ], "status": "success", "winner": true, "rank": 1 }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "score": 9.0, "judge_mean": 9.0, "judge_passes": [ 9.55, 9.15, 8.3 ], "status": "success", "winner": false, "rank": 2 }, { "model": "mimo-v2.6-flash", "model_id": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "score": 8.95, "judge_mean": 8.95, "judge_passes": [ 8.7, 9.05, 9.1 ], "status": "success", "winner": false, "rank": 3 }, { "model": "muse-spark-1.3-contributor", "model_id": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "score": 8.65, "judge_mean": 8.65, "judge_passes": [ 8.45, 8.6, 8.9 ], "status": "success", "winner": false, "rank": 4 }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4-flash", "provider": "commandcode", "family": "deepseek", "score": 8.57, "judge_mean": 8.57, "judge_passes": [ 8.65, 8.9, 8.15 ], "status": "success", "winner": false, "rank": 5 }, { "model": "hy3", "model_id": "hy3", "provider": "opencode-go", "family": "tencent", "score": 8.48, "judge_mean": 8.48, "judge_passes": [ 9.5, 8.7, 7.25 ], "status": "success", "winner": false, "rank": 6 }, { "model": "longcat-2.0", "model_id": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "score": 8.35, "judge_mean": 8.35, "judge_passes": [ 8.35, 8.15, 8.55 ], "status": "success", "winner": false, "rank": 7 }, { "model": "ling-3.0-flash-fin-free", "model_id": "ling-3.0-flash-fin-free", "provider": "commandcode", "family": "ling", "score": 7.18, "judge_mean": 7.18, "judge_passes": [ 7.6, 6.0, 7.95 ], "status": "success", "winner": false, "rank": 8 }, { "model": "minimax-m3", "model_id": "minimax-m3", "provider": "opencode-go", "family": "minimax", "score": 0.0, "status": "dnf", "winner": false, "rank": 0, "error": "both hops failed: opencode-go returned an invalid JSON response and the commandcode rescue minimaxai/minimax-m3 is plan-gated on this account (403; bare slug -> 'Model/provider not recognized')" }, { "model": "space-bunny-free", "model_id": "space-bunny-free", "provider": "opencode-go", "family": "space", "score": 0.0, "status": "dnf", "winner": false, "rank": 0, "error": "both hops failed: opencode-go returned an invalid JSON response and the commandcode rescue stealth/space-bunny-alpha returned an empty stream (no text-delta events) across 3 attempts" }, { "model": "pixel-canary", "model_id": "pixel-canary", "provider": "commandcode", "family": "pixel", "score": 0.0, "status": "dnf", "winner": false, "rank": 0, "error": "RETIRED UPSTREAM 2026-10-01: CommandCode returned 'Pixel Canary is no longer available. The stealth preview ended on September 30th'; the opencode-go rescue reported the model unavailable" } ] }, { "match_id": "arena-production-voice-agents-architecture-barge-in", "date": "2026-10-02", "type": "guide", "topic": "Production voice agents: architecture, transport (WebRTC vs WebSocket vs telephony), cascaded STT+LLM+TTS vs speech-to-speech, VAD/turn detection, barge-in handling, latency budget, and the cost of interrupted synthesis.", "blog": "niteagent", "slug": "production-voice-agents-architecture-barge-in", "winner": "longcat-2.0", "plan_model": "deepseek-v4-pro", "research_model": "mimo-v2.6-pro", "qa_model": "claude-sonnet-5-5 (commandcode)", "threshold_met": true, "results": [ { "model": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "status": "success", "score": 8.9, "judge_passes": [ 9.15, 9.15, 8.4 ], "note": "WINNER. Judge multipass mean 8.90. 2,642w draft; QA-cleaned to 2,504w.", "model_id": "longcat-2.0" }, { "model": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "status": "success", "score": 8.8, "judge_passes": [ 8.7, 8.8, 8.9 ], "note": "PASS. Judge mean 8.80.", "model_id": "mimo-v2.6-flash" }, { "model": "minimax-m3", "provider": "opencode-go", "family": "minimax", "status": "success", "score": 8.69, "judge_passes": [ 8.85, 8.93, 8.3 ], "note": "PASS. Judge mean 8.69.", "model_id": "minimax-m3" }, { "model": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "status": "success", "score": 8.67, "judge_passes": [ 8.8, 8.8, 8.4 ], "note": "PASS. Judge mean 8.67.", "model_id": "glm-5.3-flash" }, { "model": "hy3", "provider": "opencode-go", "family": "tencent", "status": "success", "score": 8.58, "judge_passes": [ 8.75, 8.5, 8.5 ], "note": "PASS. Judge mean 8.58.", "model_id": "hy3" }, { "model": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "status": "success", "score": 8.5, "judge_passes": [ 8.65, 8.55, 8.3 ], "note": "PASS. First draft was short and missing the HERO_IMAGE_PROMPT line; re-run once per the hero-prompt gate (3,577w), then judged. Judge mean 8.50.", "model_id": "deepseek-v4-flash" }, { "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "status": "success", "score": 8.47, "judge_passes": [ 8.6, 8.2, 8.6 ], "note": "PASS. Judge mean 8.47.", "model_id": "muse-spark-1.3-contributor" }, { "model": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "status": "success", "score": 8.15, "judge_passes": [ 8.0, 8.15, 8.3 ], "note": "PASS. Judge mean 8.15.", "model_id": "gpt-6-luna" }, { "model": "space-bunny-free", "provider": "opencode-go", "family": "space", "status": "failed", "score": 0.0, "judge_passes": [], "note": "DNF — lane did not return before the 420s parallel-dispatch cap.", "model_id": "space-bunny-free" }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "status": "failed", "score": 0.0, "judge_passes": [], "note": "DNF — lane did not return before the 420s parallel-dispatch cap.", "model_id": "longcat-2.5-preview-free" }, { "model": "ling-3.0-flash-fin-free", "provider": "commandcode", "family": "ling", "status": "failed", "score": 0.0, "judge_passes": [], "note": "DNF — lane did not return before the 420s parallel-dispatch cap (this lane is documented as returning an empty CC stream).", "model_id": "ling-3.0-flash-fin-free" }, { "model": "ling-3.1-flash-free", "provider": "commandcode", "family": "ling", "status": "failed", "score": 0.0, "judge_passes": [], "note": "DNF — lane did not return before the 420s parallel-dispatch cap.", "model_id": "ling-3.1-flash-free" } ] }, { "match_id": "arena-credential-scoped-agents-short-lived-tokens", "date": "2026-10-03", "type": "guide", "topic": "Credential-Scoped Agents: Why API Keys Break Agent Permission Models (2026 Guide) — the agent is a workload, not a user; static API keys grant standing privilege and leak the full blast radius; migrate to short-lived, audience-bound, least-privilege tokens minted at the task boundary via workload identity federation (Anthropic WIF, OpenAI WIF, Microsoft Entra Agent ID) or OAuth token exchange (RFC 8693 + RFC 8707 audience binding + RFC 9449 DPoP), with SPIFFE/SPIRE + Vault for platform-agnostic workload identity, Okta Cross App Access as the MCP Enterprise-Managed Authorization extension, and cloud WIF (AWS OIDC/IRSA, GCP) as the subject-token source. Desk research only — no credential system was run; every cited URL returned HTTP 200 on 2026-10-03.", "blog": "niteagent", "slug": "credential-scoped-agents-short-lived-tokens", "winner": "deepseek-v4.1-flash", "plan_model": "deepseek-v4-pro", "research_model": "orchestrator in-session web passes (mimo-v2.6-pro research lane unreliable on this blog; facts measured from 30 primary-source URLs)", "qa_model": "claude-sonnet-5-5", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "passes": [ 8.35, 8.55, 7.85 ], "mean": 8.25, "score": 8.25, "word_count": 2986, "status": "ok", "winner": true, "note": "PASS. Only draft inside the mean>=8.0 gate. Winner after QA fixes applied (scope-format consistency, uncited stat removed, tense/overclaim fixes, third internal link added).", "model_id": "deepseek-v4-flash" }, { "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "passes": [ 8.6, 6.65 ], "mean": 7.625, "score": 7.625, "word_count": 2496, "status": "ok", "note": "Only draft with every H2 lead inside the 40-60 word band; 2 parseable judge passes.", "model_id": "muse-spark-1.3-contributor" }, { "model": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "passes": [ 6.85, 7.85, 7.35 ], "mean": 7.35, "score": 7.35, "word_count": 3172, "status": "ok", "model_id": "gpt-6-luna" }, { "model": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "passes": [ 6.7, 7.5, 5.7 ], "mean": 6.633, "score": 6.633, "word_count": 2578, "status": "ok", "note": "H2 lead paragraphs far outside the 40-60 word contract (several >100 words).", "model_id": "glm-5.3-flash" }, { "model": "hy3", "provider": "opencode-go", "family": "hy3", "passes": [ 5.5, 7.85, 5.35 ], "mean": 6.233, "score": 6.233, "word_count": 2231, "status": "ok", "note": "H2 lead paragraphs far outside the 40-60 word contract; highest judge noise (spread 2.5).", "model_id": "hy3" }, { "model": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "passes": [], "mean": 0.0, "score": 0.0, "word_count": 0, "status": "dnf", "note": "DNF — builder pool hit the 420s tool cap before this lane returned.", "model_id": "mimo-v2.6-flash" }, { "model": "space-bunny-free", "provider": "opencode-go", "family": "space", "passes": [], "mean": 0.0, "score": 0.0, "word_count": 0, "status": "dnf", "note": "DNF — builder pool hit the 420s tool cap (documented hang risk on this lane).", "model_id": "space-bunny-free" }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "passes": [], "mean": 0.0, "score": 0.0, "word_count": 0, "status": "dnf", "note": "DNF — builder pool hit the 420s tool cap before this lane returned.", "model_id": "longcat-2.5-preview-free" } ] }, { "match_id": "arena-filesystem-based-agent-memory-the-default-nobody-studied", "date": "2026-10-04", "type": "deep-dive", "topic": "Filesystem-Based Agent Memory: The Default Nobody Studied — a desk-research breakdown of arXiv 2607.26637 (Zhou et al., 2026, 59 pages): long-term agent memory is already a directory tree of markdown files (SKILL.md, MEMORY.md, NOTES.md, CLAUDE.md, AGENTS.md, hosted /memories dirs), and this is the first systematic study of that default. Three roles around one store (management agent organizes; search agent answers with cited sources; execution agent distills trajectories into skills); organization buys search economy (organized stores roughly halve retrieval cost where material is large) but NOT accuracy; organization erodes for all but the strongest management agent; and the toolset alone reshapes the store as strongly as swapping the model. Read alongside shipping tooling: Anthropic's just-in-time context engineering and six-operation memory tool, Letta's git-backed MemFS, Basic Memory, Mem0, Zep and Hermes context files. Desk research only — no hands-on benchmark; 33 candidate primary-source URLs verified HTTP 200 on 2026-10-04. 10-lane builder array dispatched; 7 drafts returned, 3 DNF (ling-3.0-flash-sante empty CC stream; space-bunny-free and longcat-2.5-preview-free timed out). All 7 drafts judged with qwen3.7-flash multipass (3 passes, mean-gated) behind a measurement-carrying verified-fact preamble; winner glm-5.3-flash 8.93 (threshold met).", "blog": "niteagent", "slug": "filesystem-based-agent-memory-the-default-nobody-studied", "winner": "glm-5.3-flash", "plan_model": "deepseek-v4-pro", "research_model": "orchestrator in-session web passes (mimo-v2.6-pro/commandcode research lane unreliable on this VM; paper + 33 primary-source URLs fetched and status-verified 2026-10-04)", "qa_model": "claude-sonnet-5-5", "threshold_met": true, "results": [ { "model": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "passes": [ 8.95, 8.7, 9.15 ], "mean": 8.93, "score": 8.93, "word_count": 2474, "status": "ok", "winner": true, "note": "PASS. Highest mean, tightest spread (0.45). Only draft clearing the gate on every pass; 52 inline external links, 7 internal. Winner after QA fixes (table contradiction + unsourced-cell labelling, LongMemEval mis-analogy, novelty overclaims softened, filler cut).", "model_id": "glm-5.3-flash" }, { "model": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "passes": [ 8.55, 9.25, 7.55 ], "mean": 8.45, "score": 8.45, "word_count": 2591, "status": "ok", "model_id": "deepseek-v4-flash", "note": "PASS. 51 external links. Noisy spread 1.70 (one 3.00 quality pass)." }, { "model": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "passes": [ 9.25, 8.85, 7.2 ], "mean": 8.43, "score": 8.43, "word_count": 2320, "status": "ok", "note": "PASS. 45 external links. Noisy spread 2.05 (one 4.00 quality pass).", "model_id": "mimo-v2.6-flash" }, { "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "passes": [ 8.5, 7.1, 8.5 ], "mean": 8.03, "score": 8.03, "word_count": 1908, "status": "ok", "note": "PASS by a hair; 51 external links.", "model_id": "muse-spark-1.3-contributor" }, { "model": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "passes": [ 7.55, 7.95, 8.5 ], "mean": 8.0, "score": 8.0, "word_count": 2645, "status": "ok", "note": "PASS exactly at the 8.0 boundary; 33 external links.", "model_id": "gpt-6-luna" }, { "model": "poolside-laguna-s-2.1", "provider": "commandcode", "family": "poolside", "passes": [ 5.55, 6.2, 5.25 ], "mean": 5.67, "score": 5.67, "word_count": 3041, "status": "ok", "model_id": "poolside-laguna-s-2.1", "note": "FAIL. Wrote zero inline citations (Sourcing 1-2/10) — no links at all." }, { "model": "hy3", "provider": "opencode-go", "family": "hy3", "passes": [ 4.65, 4.65, 4.55 ], "mean": 4.62, "score": 4.62, "word_count": 1929, "status": "ok", "note": "FAIL. Used bare-URL brackets [https://url] (66x) instead of inline markdown links — Sourcing 1-2/10.", "model_id": "hy3" }, { "model": "ling-3.0-flash-sante", "provider": "commandcode", "family": "ling", "passes": [], "mean": 0.0, "score": 0.0, "word_count": 0, "status": "failed", "note": "DNF — empty commandcode stream (0 text deltas).", "model_id": "ling-3.0-flash-sante" }, { "model": "space-bunny-free", "provider": "opencode-go", "family": "space", "passes": [], "mean": 0.0, "score": 0.0, "word_count": 0, "status": "failed", "note": "DNF — timed out (no draft within the lane cap).", "model_id": "space-bunny-free" }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "passes": [], "mean": 0.0, "score": 0.0, "word_count": 0, "status": "failed", "note": "DNF — timed out (no draft within the lane cap).", "model_id": "longcat-2.5-preview-free" } ] }, { "match_id": "arena-agent-skills-skillmd-authoring-portability-security", "date": "2026-10-05", "type": "guide", "topic": "Agent Skills in 2026: the SKILL.md authoring, cross-harness portability, and supply-chain security guide. Desk research only — every figure was read from primary sources between 2026-10-05 (spec/vendor docs, GitHub API star counts) and was not hands-on tested. Covers the Anthropic Oct 16, 2025 introduction and the Dec 18, 2025 open-standard publication at agentskills.io; the SKILL.md directory contract (SKILL.md required; optional scripts/, references/, assets/); the six frontmatter fields (name max 64 chars lowercase/hyphen rules matching the parent dir, description max 1,024 chars, optional license/compatibility<=500/metadata, experimental allowed-tools); three-level progressive disclosure (~100 tokens metadata at startup, full body under 5,000 tokens on activation, resources on demand; keep SKILL.md under 500 lines); the 'skills-ref validate ./my-skill' checker that validates frontmatter only; Microsoft Agent Framework's four-stage variant (Advertise, Load, Read resources, Run scripts); the client roster (Hermes Agent, Claude, Cursor, GitHub Copilot, Goose, OpenHands, Amp, Tabnine, Roo Code, JetBrains Junie, Qodo, Spring AI, Snowflake Cortex Code, Pulumi Neo, OpenClaw, Command Code); GitHub API star counts on 2026-10-05 (obra/superpowers 295,590; anthropics/skills 179,764; addyosmani/agent-skills 101,480; wshobson/agents 40,216; agentskills/agentskills 25,924; microsoft/agent-skills 3,081; modelcontextprotocol/servers 91,017); the Snyk ToxicSkills audit (Feb 5, 2026: 3,984 skills scanned, 13.4% critical, 36.82% with a flaw, 76 malicious payloads, no signing/review/sandbox); and arXiv 2510.26328 on trivially simple skill prompt injections and 'Don't ask again' approval carry-over. Plan contract (deepseek-v4-pro via commandcode) with verbatim MUST include / MUST NOT include lists copied into every brief. 10-lane canonical array: deepseek-v4.1-flash DNF (16K output cap consumed by reasoning on both attempts) and ling-3.0-flash-sante DNF (commandcode empty stream); the 8 survivors were judged with qwen3.7-flash multipass (3 passes, mean-gated) behind a fact-carrying preamble; all 8 cleared 8.0; winner muse-spark-1.3-contributor 9.63 (space-bunny 9.57, glm 9.49, mimo 9.43, longcat 9.40, poolside 9.33, luna 8.83, hy3 8.57). QA (claude-sonnet-5-5) returned FAIL with 13 advisory defects; the winner's post shipped with the surgical fixes applied (clients anchor, star-count phrasing, peer-reviewed->published, least-privilege closing, max-vs-under wording, allowed-tools marked advisory). All 16 external URLs and the hero verified HTTP 200.", "blog": "niteagent", "slug": "agent-skills-skillmd-authoring-portability-security", "winner": "muse-spark-1.3-contributor", "plan_model": "deepseek-v4-pro", "research_model": "orchestrator-web-pass (in-session primary-source fetches + GitHub API)", "qa_model": "claude-sonnet-5-5", "threshold_met": true, "results": [ { "model": "muse-spark-1.3-contributor", "model_id": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "score": 9.63, "mean": 9.63, "passes": [ { "accuracy": 10.0, "sourcing": 10.0, "quality": 9.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.7 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.5 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 9.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.7 } ], "status": "ok", "word_count": 2238 }, { "model": "space-bunny-free", "model_id": "space-bunny-free", "provider": "opencode-go", "family": "space", "score": 9.57, "mean": 9.57, "passes": [ { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.5 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.5 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 9.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.7 } ], "status": "ok", "word_count": 2394 }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "score": 9.49, "mean": 9.49, "passes": [ { "accuracy": 10.0, "sourcing": 10.0, "quality": 9.0, "practicality": 10.0, "engagement": 10.0, "weighted": 9.8 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 9.0, "engagement": 9.0, "weighted": 9.35 }, { "accuracy": 9.5, "sourcing": 10.0, "quality": 8.5, "practicality": 9.5, "engagement": 8.5, "weighted": 9.33 } ], "status": "ok", "word_count": 2259 }, { "model": "mimo-v2.6-flash", "model_id": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "score": 9.43, "mean": 9.43, "passes": [ { "accuracy": 10.0, "sourcing": 10.0, "quality": 9.0, "practicality": 9.0, "engagement": 9.0, "weighted": 9.55 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 10.0, "engagement": 8.0, "weighted": 9.4 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 9.0, "engagement": 9.0, "weighted": 9.35 } ], "status": "ok", "word_count": 2737 }, { "model": "longcat-2.5-preview-free", "model_id": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "score": 9.4, "mean": 9.4, "passes": [ { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.5 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 9.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.7 }, { "accuracy": 9.0, "sourcing": 10.0, "quality": 7.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.0 } ], "status": "ok", "word_count": 2431 }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "commandcode", "family": "poolside", "score": 9.33, "mean": 9.33, "passes": [ { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 10.0, "engagement": 9.0, "weighted": 9.5 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 7.0, "practicality": 9.0, "engagement": 9.0, "weighted": 9.15 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 8.0, "practicality": 9.0, "engagement": 9.0, "weighted": 9.35 } ], "status": "ok", "word_count": 1654 }, { "model": "gpt-6-luna", "model_id": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "score": 8.83, "mean": 8.83, "passes": [ { "accuracy": 10.0, "sourcing": 10.0, "quality": 7.0, "practicality": 9.5, "engagement": 9.0, "weighted": 9.22 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 4.0, "practicality": 9.0, "engagement": 8.0, "weighted": 8.45 }, { "accuracy": 10.0, "sourcing": 10.0, "quality": 5.5, "practicality": 9.0, "engagement": 8.5, "weighted": 8.8 } ], "status": "ok", "word_count": 2651 }, { "model": "hy3", "model_id": "hy3", "provider": "opencode-go", "family": "hy3", "score": 8.57, "mean": 8.57, "passes": [ { "accuracy": 9.0, "sourcing": 9.0, "quality": 7.0, "practicality": 9.0, "engagement": 8.0, "weighted": 8.5 }, { "accuracy": 9.0, "sourcing": 9.0, "quality": 7.0, "practicality": 9.0, "engagement": 8.0, "weighted": 8.5 }, { "accuracy": 9.0, "sourcing": 9.0, "quality": 8.0, "practicality": 9.0, "engagement": 8.0, "weighted": 8.7 } ], "status": "ok", "word_count": 1957 }, { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "score": 0.0, "mean": 0.0, "passes": [], "triaged_skipped": true, "status": "failed", "note": "truncated at 835 words on the parallel dispatch and again at 42 words on retry (16,384-token output cap consumed by reasoning); no HERO_IMAGE_PROMPT line; JEV band=reject" }, { "model": "ling-3.0-flash-sante", "model_id": "ling-3.0-flash-sante", "provider": "commandcode", "family": "ling", "score": 0.0, "mean": 0.0, "passes": [], "status": "dnf", "note": "DNF — commandcode empty stream (chronic for the ling free lanes)" } ] }, { "match_id": "arena-stop-runaway-ai-agents-kill-switch-cancellation-spend-limits", "date": "2026-10-06", "type": "guide", "topic": "Stopping a runaway AI agent in production: cancellation semantics (Inngest, Temporal, Cloudflare Workflows, LangGraph, Restate), timeouts, and spend-side kill switches (OpenAI, LiteLLM, OpenRouter). Fills roadmap Phase 12 (Cancellation & kill switches, Lifecycle observability) and Phase 8 (Agent cost control).", "blog": "niteagent", "slug": "stop-runaway-ai-agents-kill-switch-cancellation-spend-limits", "winner": "mimo-v2.6-flash", "plan_model": "deepseek-v4-pro", "research_model": "mimo-v2.6-pro (research lane stalled/killed; facts measured in-session by orchestrator web passes + GitHub API on 2026-10-06)", "qa_model": "claude-sonnet-5-5", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "status": "success", "score": 9.633, "per_pass": [ 9.7, 9.7, 9.5 ], "mean": 9.633, "word_count": 3822, "winner": false, "model_id": "deepseek-v4.1-flash" }, { "model": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "status": "success", "score": 9.967, "per_pass": [ 9.9, 10.0, 10.0 ], "mean": 9.967, "word_count": 3980, "winner": true, "model_id": "mimo-v2.6-flash" }, { "model": "glm-5.3-flash", "provider": "opencode-go", "family": "glm", "status": "success", "score": 9.467, "per_pass": [ 9.7, 9.7, 9.0 ], "mean": 9.467, "word_count": 3008, "winner": false, "model_id": "glm-5.3-flash" }, { "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "status": "success", "score": 9.667, "per_pass": [ 9.7, 9.7, 9.6 ], "mean": 9.667, "word_count": 3193, "winner": false, "model_id": "muse-spark-1.3-contributor" }, { "model": "hy3", "provider": "opencode-go", "family": "hy3", "status": "success", "score": 9.517, "per_pass": [ 9.7, 9.7, 9.15 ], "mean": 9.517, "word_count": 2317, "winner": false, "model_id": "hy3" }, { "model": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "status": "success", "score": 9.45, "per_pass": [ 9.6, 9.15, 9.6 ], "mean": 9.45, "word_count": 2885, "winner": false, "model_id": "gpt-6-luna" }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "status": "success", "score": 9.45, "per_pass": [ 8.95, 9.7, 9.7 ], "mean": 9.45, "word_count": 3433, "winner": false, "model_id": "longcat-2.5-preview-free" }, { "model": "poolside-laguna-s-2.1", "provider": "commandcode", "family": "poolside", "status": "success", "score": 8.217, "per_pass": [ 8.25, 8.35, 8.05 ], "mean": 8.217, "word_count": 2573, "winner": false, "model_id": "poolside-laguna-s-2.1" }, { "model": "space-bunny-free", "provider": "opencode-go", "family": "space", "status": "failed", "score": 0.0, "mean": 0.0, "word_count": 0, "error": "DNF (lane failed / no draft)", "model_id": "space-bunny-free" }, { "model": "ling-3.0-flash-sante", "provider": "commandcode", "family": "ling", "status": "failed", "score": 0.0, "mean": 0.0, "word_count": 0, "error": "DNF (lane failed / no draft)", "model_id": "ling-3.0-flash-sante" } ] }, { "match_id": "arena-agent-rate-limit-governor-build-log", "date": "2026-10-07", "type": "build-log", "topic": "Build log: a client-side rate-limit governor for production AI agents - token-bucket admission control across separate RPM/TPM/RPD/concurrency axes, spend-cap 429 classification (enforced_spend_limit_reached, never retryable), full-jitter backoff, distributed limits and Retry-After edge cases. Fills roadmap Phase 5 (Rate limit management) and Phase 8 (cost monitoring).", "blog": "niteagent", "slug": "agent-rate-limit-governor-build-log", "winner": "glm-5.3-flash", "plan_model": "deepseek-v4-pro", "research_model": "mimo-v2.6-pro (commandcode research lane, 1341-word dossier 2026-10-07; facts cross-verified in-session by orchestrator web passes over 13 primary-source URLs)", "qa_model": "claude-sonnet-5-5", "threshold_met": true, "results": [ { "model": "deepseek-v4.1-flash", "model_id": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "status": "success", "score": 8.65, "mean": 8.65, "per_pass": [ 8.6, 8.65, 8.7 ], "word_count": 2419, "winner": false }, { "model": "mimo-v2.6-flash", "model_id": "mimo-v2.6-flash", "provider": "opencode-go", "family": "mimo", "status": "success", "score": 8.92, "mean": 8.92, "per_pass": [ 9.3, 8.5, 8.95 ], "word_count": 2805, "winner": false }, { "model": "glm-5.3-flash", "model_id": "glm-5.3-flash", "provider": "zai", "family": "glm", "status": "success", "score": 8.78, "mean": 8.78, "per_pass": [ 9.15, 8.6, 8.6 ], "word_count": 2067, "winner": true }, { "model": "muse-spark-1.3-contributor", "model_id": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "muse", "status": "success", "score": 8.53, "mean": 8.53, "per_pass": [ 8.6, 8.3, 8.7 ], "word_count": 2963, "winner": false }, { "model": "hy3", "model_id": "hy3", "provider": "opencode-go", "family": "hy3", "status": "success", "score": 8.17, "mean": 8.17, "per_pass": [ 8.0, 7.9, 8.6 ], "word_count": 2113, "winner": false }, { "model": "gpt-6-luna", "model_id": "gpt-6-luna", "provider": "commandcode", "family": "gpt", "status": "success", "score": 8.62, "mean": 8.62, "per_pass": [ 9.15, 8.3, 8.4 ], "word_count": 2539, "winner": false }, { "model": "longcat-2.5-preview-free", "model_id": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "status": "success", "score": 8.67, "mean": 8.67, "per_pass": [ 8.6, 8.7, 8.7 ], "word_count": 2615, "winner": false }, { "model": "ling-3.0-flash-sante", "model_id": "ling-3.0-flash-sante", "provider": "commandcode", "family": "ling", "status": "success", "score": 8.32, "mean": 8.32, "per_pass": [ 8.1, 8.25, 8.6 ], "word_count": 2125, "winner": false }, { "model": "poolside-laguna-s-2.1", "model_id": "poolside-laguna-s-2.1", "provider": "commandcode", "family": "poolside", "status": "success", "score": 6.75, "mean": 6.75, "per_pass": [ 6.75, 6.55, 6.95 ], "word_count": 2418, "winner": false }, { "model": "space-bunny-free", "model_id": "space-bunny-free", "provider": "opencode-go", "family": "space", "status": "failed", "score": 0.0, "mean": 0.0, "per_pass": [], "word_count": 0, "winner": null } ] }, { "id": 90, "match_id": "na-home-20261007", "date": "2026-10-07", "type": "homepage_redesign_spec", "topic": "NiteAgent homepage redesign for AdSense 'high value site' goal, built from updated DESIGN.md (emilkowalski/skills + pbakaus/impeccable + leonxlnx/taste-skill + Figma MCP catalog + microsoft/playwright). 9-lane canonical roster spec arena.", "winner": "muse-spark-1.3-contributor", "threshold_met": true, "plan_model": null, "research_model": "DESIGN.md v2 (25099 bytes) mined from 5 sources by glm-5.3-flash subagent; embedded in brief", "judge_rubric": "c1-c5 canonical (data 30 / layout 25 / fidelity 20 / code 15 / ux 10)", "judge_model": "deepseek-v4-pro via openrouter, 3-pass multipass, rotated builder order", "judges": 3, "results": [ { "rank": 1, "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "score": 8.13, "passes": [ 8.2, 7.8, 8.4 ] }, { "rank": 2, "model": "deepseek-v4.1-flash", "provider": "commandcode", "score": 7.93, "passes": [ 7.4, 8.9, 7.5 ] }, { "rank": 3, "model": "mimo-v2.6-flash", "provider": "opencode-go", "score": 7.67, "passes": [ 7.2, 8.7, 7.1 ] }, { "rank": 4, "model": "hy3", "provider": "opencode-go", "score": 7.7, "passes": [ 7.4, 8.2, 7.5 ] }, { "rank": 5, "model": "gpt-6-luna", "provider": "commandcode", "score": 7.7, "passes": [ 6.7, 8.9, 7.5 ] }, { "rank": 6, "model": "poolside/laguna-s-2.1-free", "provider": "commandcode", "score": 7.17, "passes": [ 6.8, 8.0, 6.7 ] } ], "dnf": [ { "model": "glm-5.3-flash", "provider": "zai", "reason": "timeout after 2 retries" }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "reason": "timeout after 2 retries" }, { "model": "ling-3.0-flash-sante", "provider": "commandcode", "reason": "empty stream (known failure mode)" } ], "lanes_declared": 9, "drafts": 6, "judge_caveats": "Pass-to-pass spread was large (gpt-6-luna 6.7 to 8.9). Passes 1/3 penalized all builders on an em-dash criterion the judge prompt itself introduced, contradicting the brief. muse-spark won 2/3 passes and leads on the discriminating pass 2 evidence. Winner selected at 8.13 mean; runner-up deepseek-v4.1-flash at 7.93 considered for its 221-card grid but rejected: brief favored dense table for the archive.", "shipped": { "spec_deviations": [ "clip-path accordion reveal dropped (DESIGN.md motion contract: transform/opacity only)", "missing-value placeholder: word None, not em-dash", "Show-all toggle dropped (filtered views show all; load-more returns on clear)", "token names mapped to real global.css names (--bg-card, --text-primary, --border, --accent)" ], "gates": { "layout_audit_local": "PASS (js_errors, parity d0, cols 1/2/4/4/6/6, full rows, 1 control group, imagery, load-more)", "verify_built_links": "ALL LINKS RESOLVE (run from dist/ root; the script's dirname derivation breaks on root-level pages)", "functional_test": "9/9 (count, load-more 40/80, filter 49, search 15, empty state, clear, sort, accordion, zero JS errors)", "vision_pass": "orchestrator interactive (glm-5.3-flash): 3 screenshots reviewed; no crushed rows, no dead zones, widescreen uses width (Gate column visible, 6-col grid, side-by-side panels)", "layout_audit_live": "PASS after cache-bust (first 1280px fetch hit stale edge cache serving the old page)" }, "commit": "6806d27", "deploy": "d281f0de.crypto-nite-blog.pages.dev -> niteagent.com, verified live" } } ], "leaderboard": [ { "model": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "matches": 88, "wins": 42, "losses": 46, "avgScore": 8.45, "bestScore": 9.8, "rank": 1 }, { "model": "poolside-laguna-s-2.1", "provider": "openrouter", "family": "poolside", "matches": 67, "wins": 11, "losses": 56, "avgScore": 8.3, "bestScore": 9.72, "rank": 2 }, { "model": "glm-5.3-flash", "provider": "opencode-go", "family": "z-ai", "matches": 26, "wins": 9, "losses": 17, "avgScore": 8.86, "bestScore": 9.73, "rank": 3 }, { "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "meta", "matches": 14, "wins": 3, "losses": 11, "avgScore": 8.68, "bestScore": 9.67, "rank": 4 }, { "model": "mimo-v2.6-flash", "provider": "opencode-go", "family": "xiaomi", "matches": 11, "wins": 2, "losses": 9, "avgScore": 9.01, "bestScore": 9.97, "rank": 5 }, { "model": "gpt-6-luna", "provider": "commandcode", "family": "openai", "matches": 11, "wins": 1, "losses": 10, "avgScore": 8.54, "bestScore": 9.72, "rank": 6 }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "matches": 7, "wins": 0, "losses": 7, "avgScore": 9.15, "bestScore": 9.45, "rank": 7 }, { "model": "minimax-m3", "provider": "cline-pass", "family": "minimax", "matches": 5, "wins": 0, "losses": 5, "avgScore": 8.79, "bestScore": 9.8, "rank": 8 }, { "model": "ling-3.0-flash-sante", "provider": "commandcode", "family": "ling", "matches": 4, "wins": 0, "losses": 4, "avgScore": 8.32, "bestScore": 8.32, "rank": 9 }, { "model": "hy3", "provider": "opencode-go", "family": "tencent", "matches": 13, "wins": 0, "losses": 13, "avgScore": 8.23, "bestScore": 9.52, "rank": 10 } ], "model_pool": [ { "model": "deepseek-v4.1-flash", "provider": "commandcode", "family": "deepseek", "role": "builder", "status": "active", "matches_played": 48, "benchmark_score": 8.3, "benchmark_weight": 0.83, "reliability": 0.6, "fallback": "opencode-go bare deepseek-v4-flash" }, { "model": "mimo-v2.6-flash", "provider": "opencode-go", "family": "xiaomi", "role": "builder", "status": "active", "context_length": 1048576, "fallback": "commandcode xiaomi/mimo-v2.6-flash" }, { "model": "glm-5.3-flash", "provider": "opencode-go", "family": "z-ai", "role": "builder", "status": "active", "context_length": 1310720, "fallback": "openrouter z-ai/glm-5.3-flash" }, { "model": "muse-spark-1.3-contributor", "provider": "opencode-go", "family": "meta", "role": "builder", "status": "active", "matches_played": 1, "fallback": "commandcode meta/muse-spark-1.3-contributor" }, { "model": "minimax-m3", "provider": "cline-pass", "family": "minimax", "role": "builder", "status": "active", "fallback": "commandcode bare minimax-m3 (trace-strip reasoning content)", "retired_reason": "2026-10-02 user directive: too expensive for intelligence tier as builder" }, { "model": "hy3", "provider": "opencode-go", "family": "tencent", "role": "builder", "status": "active", "fallback": "commandcode tencent/hy3-paid", "note": "restored 2026-10-02 (same-day reversal of retirement)" }, { "model": "gpt-6-luna", "provider": "commandcode", "family": "openai", "role": "builder", "status": "active", "fallback": "opencode-go bare gpt-6-luna | openrouter openai/gpt-6-luna" }, { "model": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "role": "builder", "status": "retired", "fallback": "none (CommandCode plan-blocked)", "reason": "superseded by longcat-2.5-preview-free 2026-10-01", "retired_date": "2026-10-01" }, { "model": "space-bunny-free", "provider": "opencode-go", "family": "space", "role": "builder", "status": "retired", "fallback": "commandcode stealth/space-bunny-alpha (1M ctx, Stealth route, probe-OK)" }, { "model": "deepseek-v4-pro", "provider": "commandcode", "role": "planner", "status": "active", "family": "deepseek", "matches_played": 1 }, { "model": "qwen3.7-flash", "model_id": "qwen3.7-flash", "role": "judge", "provider": "openrouter", "context_length": 1000000, "status": "active" }, { "model": "mimo-v2.6-pro", "provider": "commandcode", "role": "research", "note": "last 2 matches ran research on deepseek-v4-pro (2026-09-30)", "status": "active" }, { "model": "ling-3.0-flash-fin-free", "provider": "commandcode", "family": "ling", "role": "builder", "status": "retired", "context_length": 262144, "fallback": "opencode-go ling-3.0-flash-fin-free (not serving 2026-09-25 - Model is unavailable.; CC-only lane, wire inclusionai/ling-3.0-flash-sante:free, FREE deal)", "retired_reason": "2026-10-02 user directive: chronic CC empty-stream (DNF in match-036)" }, { "model": "longcat-2.5-preview-free", "provider": "opencode-go", "family": "longcat25", "role": "builder", "status": "active", "context_length": 262144 }, { "model": "ling-3.1-flash-free", "provider": "commandcode", "family": "ling", "role": "builder", "status": "retired", "context_length": 262144, "fallback": "CC-only lane (og does not serve it; OR has only 3.0 variants). Wire id inclusionai/ling-3.1-flash:free. Reasoning-model: needs max_tokens>=2000 (REASONING_MODELS member).", "retired_reason": "2026-10-02 user directive: chronic CC empty-stream (DNF in match-036)" }, { "model": "claude-sonnet-5-5", "provider": "commandcode", "family": "claude", "role": "qa", "status": "active", "fallback": "glm-5.3 (opencode-go) -> glm-5.3-flash (zai), pinned in qa-glm53.py (2026-10-01 upgrade)" }, { "model": "ling-3.0-flash-sante", "provider": "commandcode", "family": "ling", "role": "builder", "status": "active" }, { "model": "claude-haiku-5-5", "provider": "claude-cli", "family": "claude", "role": "builder", "status": "active", "matches_played": 0, "fallback": "commandcode claude-haiku-5-5" }, { "model": "qwen3.8-max", "provider": "cline-pass", "family": "qwen", "role": "builder", "status": "active", "matches_played": 0, "fallback": "" }, { "model": "poolside/laguna-s-2.1-free", "role": "builder", "status": "retired", "provider": "commandcode", "family": "poolside", "reason": "lane removed — it ran on PAID OpenRouter", "retired_date": "2026-10-16" } ], "stability": { "last_updated": "2026-10-08", "model_stats": { "deepseek-v4-flash": { "matches": 80, "wins": 43, "losses": 37, "avg_score": 8.4, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "deepseek-v4-preview": { "total_calls": 25, "successes": 15, "availability": 60, "reliability": 0.6, "status": "error" }, "deepseek-v4": { "total_calls": 10, "successes": 10, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen-3.7-max": { "total_calls": 28, "successes": 23, "availability": 82, "reliability": 0.2, "status": "fail" }, "qwen-3.6-plus": { "total_calls": 5, "successes": 0, "availability": 0, "reliability": 0.0, "status": "fail" }, "xiaomi-mimo-2.5": { "total_calls": 18, "successes": 14, "availability": 78, "reliability": 0.0, "status": "fail" }, "mimo-v2.5": { "matches": 67, "wins": 16, "losses": 51, "avg_score": 8.14, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "gpt-oss-120b": { "total_calls": 35, "successes": 33, "availability": 94, "reliability": 0.0, "status": "fail" }, "gemma-4-31b": { "total_calls": 15, "successes": 14, "availability": 93, "reliability": 0.0, "status": "fail" }, "mistral-large-latest": { "total_calls": 45, "successes": 43, "availability": 96, "reliability": 0.5, "status": "pass" }, "mistral-medium-latest": { "total_calls": 10, "successes": 8, "availability": 80, "reliability": 0.0, "status": "fail" }, "mistral-small-2603": { "total_calls": 20, "successes": 18, "availability": 90, "reliability": 0.0, "status": "fail" }, "llama-3.3-70b-versatile": { "total_calls": 60, "successes": 56, "availability": 93, "reliability": 0.0, "status": "fail" }, "llama-3.1-8b-instant": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": null, "status": "untested" }, "cohere/north-mini-code:free": { "total_calls": 15, "successes": 13, "availability": 87, "reliability": 0.0, "status": "fail" }, "nvidia/nemotron-3-ultra-550b-a55b:free": { "total_calls": 10, "successes": 10, "availability": 100, "reliability": 1.0, "status": "pass" }, "nvidia/nemotron-3-super-120b-a12b:free": { "total_calls": 8, "successes": 6, "availability": 75, "reliability": 0.75, "status": "http_error" }, "glm-5.2": { "total_calls": 5, "successes": 0, "availability": 0, "reliability": 0.0, "status": "empty" }, "nousresearch/hermes-3-llama-3.1-405b:free": { "total_calls": 2, "successes": 0, "availability": 0, "reliability": 0.0, "status": "http_error" }, "stepfun/step-3.7-flash:free": { "total_calls": 0, "successes": 0, "availability": 0, "reliability": null, "status": "untested" }, "openai/gpt-oss-20b:free": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "tencent/hy3:free": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "google/gemma-4-26b-a4b-it:free": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "groq/compound-mini": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "groq/compound": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen/qwen3-32b": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "openai/gpt-oss-20b": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "meta-llama/llama-4-scout-17b-16e-instruct": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "codestral-2508": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "devstral-2512": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "magistral-medium-2509": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "mistral-medium-2604": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "ministral-14b-2512": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "open-mistral-nemo-2407": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "gpt-5.5-instant-14k": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "nexum-router": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen-3.7-max-thinking": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen-3.7-plus": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen-3.6-max-preview": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen-3.5-omni-plus": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen-3.6-plus-thinking": { "total_calls": 5, "successes": 5, "availability": 100, "reliability": 1.0, "status": "pass" }, "mistral-small-latest": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "google/gemma-4-31b-it:free": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "nvidia/nemotron-3-nano-30b-a3b:free": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "nvidia/nemotron-nano-12b-v2-vl:free": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "nvidia/nemotron-3.5-content-safety:free": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "nemotron-3-ultra-550b": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "glm-5": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "glm-5-turbo": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "glm-5.1": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "mimo-v2.5-pro": { "matches": 5, "wins": 2, "losses": 3, "avg_score": 7.94, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen-qwen3.7-flash": { "matches": 1, "wins": 0, "losses": 1, "avg_score": 6.5, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "cmd-codex": { "matches": 1, "wins": 0, "losses": 1, "avg_score": 9.23, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "openai/gpt-5.6-luna": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "qwen3.7-flash": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "glm-5.3-flash": { "total_calls": 1, "successes": 1, "availability": 1.0, "reliability": 1.0, "status": "ok", "matches": 14, "wins": 5, "losses": 9, "avg_score": 9.04 }, "z-ai/glm-5.3-flash": { "total_calls": 1, "successes": 0, "availability": 0.0, "reliability": 0.0, "status": "failed" }, "qwen/qwen3.7-flash": { "total_calls": 3, "successes": 3, "availability": 1.0, "reliability": 1.0, "status": "ok" }, "muse-spark-1.3-contributor": { "matches": 1, "wins": 1, "losses": 0, "avg_score": 9.32, "total_calls": 1, "successes": 1, "availability": 100, "reliability": 1.0, "status": "pass" }, "deepseek-v4-pro": { "matches": 0, "wins": 0, "losses": 0, "avg_score": 0, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "hy3": { "matches": 0, "wins": 0, "losses": 0, "avg_score": 0, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "longcat-2.0": { "matches": 0, "wins": 0, "losses": 0, "avg_score": 0, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "mimo-v2.6-flash": { "matches": 2, "wins": 1, "losses": 1, "avg_score": 9.38, "total_calls": 2, "successes": 2, "availability": 100.0, "reliability": 1.0, "status": "pass" }, "minimax-m3": { "matches": 2, "wins": 0, "losses": 2, "avg_score": 8.84, "total_calls": 2, "successes": 2, "availability": 100.0, "reliability": 1.0, "status": "pass" }, "gpt-6-luna": { "matches": 2, "wins": 0, "losses": 2, "avg_score": 8.71, "total_calls": 2, "successes": 2, "availability": 100.0, "reliability": 1.0, "status": "pass" }, "space-bunny-free": { "matches": 2, "wins": 0, "losses": 2, "avg_score": 8.91, "total_calls": 2, "successes": 2, "availability": 100.0, "reliability": 1.0, "status": "pass" }, "mimo-v2.6-pro": { "matches": 0, "wins": 0, "losses": 0, "avg_score": 0, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "idle" }, "ling-3.0-flash-fin-free": { "matches": 1, "wins": 0, "losses": 1, "avg_score": 0, "total_calls": 1, "successes": 0, "availability": 0.0, "reliability": 0.0, "status": "idle" }, "longcat-2.5-preview-free": { "matches": 0, "wins": 0, "losses": 0, "avg_score": 0, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "pixel-canary": { "matches": 0, "wins": 0, "losses": 0, "avg_score": 0, "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "ling-3.1-flash-free": { "matches": 1, "wins": 0, "losses": 1, "avg_score": 0, "total_calls": 1, "successes": 0, "availability": 0.0, "reliability": 0.0, "status": "idle" }, "claude-sonnet-5-5": { "total_calls": 0, "successes": 0, "availability": 100, "reliability": 1.0, "status": "pass" }, "ling-3.0-flash-sante": { "matches": 1, "wins": 0, "losses": 1, "avg_score": 0, "total_calls": 1, "successes": 0, "availability": 0, "reliability": 0.0, "status": "error" }, "deepseek-v4.1-flash": { "matches": 85, "wins": 42, "losses": 43, "avg_score": 8.44, "total_calls": 85, "successes": 85, "availability": 100, "reliability": 1.0, "status": "pass" } } }, "judge_pool": [ { "model": "gemma-4-31b", "provider": "cerebras", "family": "gemma", "benchmark_score": 7.3, "benchmark_weight": 0.73, "reliability": 0.0, "matches_played": 3, "status": "pass", "role": "judge" }, { "model": "gpt-oss-120b", "provider": "cerebras", "family": "gpt", "benchmark_weight": 0.84, "benchmark_score": 8.4, "reliability": 0.0, "matches_played": 4, "status": "pass", "role": "judge" }, { "model": "llama-3.1-8b-instant", "provider": "groq", "family": "llama", "status": "pass", "role": "judge" }, { "model": "llama-3.3-70b-versatile", "provider": "groq", "family": "llama", "benchmark_weight": 0.79, "benchmark_score": 7.9, "reliability": 0.0, "matches_played": 10, "status": "pass", "role": "judge" }, { "model": "mistral-large-latest", "provider": "mistral", "family": "mistral", "benchmark_score": 8.5, "benchmark_weight": 0.85, "reliability": 0.5, "matches_played": 8, "status": "pass", "role": "judge" }, { "model": "mistral-medium-latest", "provider": "mistral", "family": "mistral", "benchmark_score": 8.0, "benchmark_weight": 0.8, "reliability": 0.0, "matches_played": 1, "status": "pass", "role": "judge" }, { "model": "mistral-small-latest", "provider": "mistral", "family": "mistral", "status": "pass", "role": "judge" }, { "model": "cohere/north-mini-code:free", "provider": "openrouter", "family": "cohere", "benchmark_score": 6.5, "benchmark_weight": 0.65, "reliability": 0.0, "matches_played": 1, "status": "pass", "role": "judge" }, { "model": "google/gemma-4-31b-it:free", "provider": "openrouter", "family": "gemma", "status": "pass", "role": "judge" }, { "model": "nvidia/nemotron-3-super-120b-a12b:free", "provider": "openrouter", "family": "nvidia", "status": "pass", "role": "judge" }, { "model": "nvidia/nemotron-3-ultra-550b-a55b:free", "provider": "openrouter", "family": "nvidia", "status": "pass", "role": "judge" }, { "model": "nvidia/nemotron-3-nano-30b-a3b:free", "provider": "openrouter", "family": "nvidia", "status": "pass", "role": "judge" }, { "model": "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free", "provider": "openrouter", "family": "nvidia", "status": "pass", "role": "judge" }, { "model": "nvidia/nemotron-nano-12b-v2-vl:free", "provider": "openrouter", "family": "nvidia", "status": "pass", "role": "judge" }, { "model": "openai/gpt-oss-20b:free", "provider": "openrouter", "family": "gpt", "status": "pass", "role": "judge" }, { "model": "google/gemma-4-26b-a4b-it:free", "provider": "openrouter", "family": "gemma", "status": "pass", "role": "judge" }, { "model": "nvidia/nemotron-3.5-content-safety:free", "provider": "openrouter", "family": "nvidia", "status": "pass", "role": "judge" }, { "model": "tencent/hy3:free", "provider": "openrouter", "family": "tencent", "status": "pass", "role": "judge" }, { "model": "stepfun/step-3.7-flash:free", "provider": "nousresearch", "family": "stepfun", "status": "pass", "role": "judge" }, { "model": "groq/compound-mini", "provider": "groq", "family": "compound", "status": "pass", "role": "judge" }, { "model": "groq/compound", "provider": "groq", "family": "compound", "status": "pass", "role": "judge" }, { "model": "qwen/qwen3-32b", "provider": "groq", "family": "qwen", "status": "pass", "role": "judge" }, { "model": "openai/gpt-oss-20b", "provider": "groq", "family": "gpt", "status": "pass", "role": "judge" }, { "model": "meta-llama/llama-4-scout-17b-16e-instruct", "provider": "groq", "family": "llama", "status": "pass", "role": "judge" }, { "model": "codestral-2508", "provider": "mistral", "family": "codestral", "benchmark_score": 8.1, "benchmark_weight": 0.81, "reliability": 0.0, "matches_played": 4, "status": "pass", "role": "judge" }, { "model": "devstral-2512", "provider": "mistral", "family": "devstral", "status": "pass", "role": "judge" }, { "model": "magistral-medium-2509", "provider": "mistral", "family": "magistral", "status": "pass", "role": "judge" }, { "model": "mistral-medium-2604", "provider": "mistral", "family": "mistral", "status": "pass", "role": "judge" }, { "model": "ministral-14b-2512", "provider": "mistral", "family": "ministral", "status": "pass", "role": "judge" }, { "model": "open-mistral-nemo-2407", "provider": "mistral", "family": "nemo", "status": "pass", "role": "judge" }, { "model": "nemotron-3-ultra-550b", "provider": "openrouter", "family": "nemotron", "benchmark_score": 8.2, "benchmark_weight": 0.82, "reliability": 0.5, "matches_played": 4, "status": "pass", "role": "judge" } ], "retired_models": [ { "model": "z-ai/glm-5.3-flash", "status": "identity-merged", "reason": "z-ai/ prefixed variant of active glm-5.3-flash; stats merged" }, { "model": "mimo-v2.5-pro", "status": "inactive", "reason": "not an active builder (judge/planner/experiment role or retired one-off)", "last_matches": 5, "retired_date": "2026-09-03" }, { "model": "cmd-codex", "status": "inactive", "reason": "not an active builder (judge/planner/experiment role or retired one-off)", "last_matches": 1, "retired_date": "2026-09-03" }, { "model": "qwen/qwen3.7-flash", "status": "inactive", "reason": "not an active builder (judge/planner/experiment role or retired one-off)", "last_matches": 1, "retired_date": "2026-09-03" }, { "model": "mimo-v2.5", "reason": "superseded by mimo-v2.6-flash 2026-09-23" }, { "model": "openai/gpt-5.6-luna", "reason": "superseded by gpt-6-luna 2026-09-23" }, { "model": "longcat-2.0", "provider": "opencode-go", "family": "longcat", "status": "retired", "reason": "removed from active lineup 2026-09-24", "retired_date": "2026-09-24" }, { "model": "openai/gpt-6-luna", "provider": "openrouter", "family": "", "status": "retired", "reason": "removed from active lineup 2026-09-24", "retired_date": "2026-09-24" }, { "model": "qwen3.8-flash", "provider": "opencode-go", "family": "qwen", "status": "retired", "reason": "removed from active lineup 2026-09-24", "retired_date": "2026-09-24" }, { "model": "pixel-canary", "provider": "commandcode", "family": "pixel", "status": "retired", "reason": "removed from active builder pool 2026-10-07 (user directive); upstream stealth preview ended", "retired_date": "2026-10-07" }, { "model": "space-bunny-free", "status": "retired", "reason": "upstream stealth preview ended 2026-10-07; lane dead upstream", "retired_date": "2026-10-07" }, { "model": "ling-3.0-flash-fin-free", "status": "retired", "reason": "user directive 2026-10-02: chronic CC empty-stream (0 text-deltas; DNF in match-036); failed heavy-call reliability gate", "retired_date": "2026-10-03" }, { "model": "ling-3.1-flash-free", "status": "retired", "reason": "user directive 2026-10-02: chronic CC empty-stream (0 text-deltas; DNF in match-036); failed heavy-call reliability gate", "retired_date": "2026-10-03" } ] }