← Back to the board

Model weights

Every model starts with an equal vote. As its calls resolve, the panel leans toward the ones whose confidence proves well-calibrated and whose recent form holds up — so a model’s weight drifts above or below the dashed equal-vote line. Nothing here is set by hand; it is read straight from the resolved log.

0.700.760.81equal votestart06-2906-3007-0207-0307-0407-0607-07anthropic:claude-opus-4-8 · start: 0.750anthropic:claude-opus-4-8 · 06-29: 0.750anthropic:claude-opus-4-8 · 06-30: 0.750anthropic:claude-opus-4-8 · 07-02: 0.750anthropic:claude-opus-4-8 · 07-03: 0.750anthropic:claude-opus-4-8 · 07-04: 0.750anthropic:claude-opus-4-8 · 07-06: 0.750anthropic:claude-opus-4-8 · 07-07: 0.760anthropic:claude-sonnet-4-6 · start: 0.750anthropic:claude-sonnet-4-6 · 06-29: 0.758anthropic:claude-sonnet-4-6 · 06-30: 0.760anthropic:claude-sonnet-4-6 · 07-02: 0.753anthropic:claude-sonnet-4-6 · 07-03: 0.753anthropic:claude-sonnet-4-6 · 07-04: 0.753anthropic:claude-sonnet-4-6 · 07-06: 0.755anthropic:claude-sonnet-4-6 · 07-07: 0.753openai:gpt-4o · start: 0.750openai:gpt-4o · 06-29: 0.771openai:gpt-4o · 06-30: 0.762openai:gpt-4o · 07-02: 0.753openai:gpt-4o · 07-03: 0.749openai:gpt-4o · 07-04: 0.749openai:gpt-4o · 07-06: 0.749openai:gpt-4o · 07-07: 0.748openrouter:deepseek/deepseek-chat · start: 0.750openrouter:deepseek/deepseek-chat · 06-29: 0.749openrouter:deepseek/deepseek-chat · 06-30: 0.740openrouter:deepseek/deepseek-chat · 07-02: 0.743openrouter:deepseek/deepseek-chat · 07-03: 0.743openrouter:deepseek/deepseek-chat · 07-04: 0.742openrouter:deepseek/deepseek-chat · 07-06: 0.750openrouter:deepseek/deepseek-chat · 07-07: 0.749openrouter:google/gemini-2.5-flash · start: 0.750openrouter:google/gemini-2.5-flash · 06-29: 0.767openrouter:google/gemini-2.5-flash · 06-30: 0.757openrouter:google/gemini-2.5-flash · 07-02: 0.750openrouter:google/gemini-2.5-flash · 07-03: 0.747openrouter:google/gemini-2.5-flash · 07-04: 0.747openrouter:google/gemini-2.5-flash · 07-06: 0.754openrouter:google/gemini-2.5-flash · 07-07: 0.751openrouter:x-ai/grok-4.3 · start: 0.750openrouter:x-ai/grok-4.3 · 06-29: 0.768openrouter:x-ai/grok-4.3 · 06-30: 0.759openrouter:x-ai/grok-4.3 · 07-02: 0.752openrouter:x-ai/grok-4.3 · 07-03: 0.748openrouter:x-ai/grok-4.3 · 07-04: 0.748openrouter:x-ai/grok-4.3 · 07-06: 0.755openrouter:x-ai/grok-4.3 · 07-07: 0.753

Current standings

ModelWeightvs equalResolved calls
anthropic:claude-opus-4-80.76+0.011
openrouter:x-ai/grok-4.30.75±0.00576
anthropic:claude-sonnet-4-60.75±0.00432
openrouter:google/gemini-2.5-flash0.75±0.00576
openrouter:deepseek/deepseek-chat0.75±0.00576
openai:gpt-4o0.75±0.00288

Weight is a model’s vote share before its own confidence is applied. A model only moves off equal once it has a genuine, sustained record — a couple of lucky calls barely shift it.