Resolves when GPT-5 is released and benchmarked on GPQA. Will resolve according to the highest compute setting the model is evaluated on (if applicable)

• Update 2025-08-12 (PST) (AI summary of creator comment): - No tools: Only tool-free evaluations of GPT-5 on GPQA will count; tool-assisted runs (e.g., retrieval, calculators, web browsing, code execution, plugins/agents) are excluded.

  • Highest compute (tool-free): If multiple tool-free evaluations exist, resolution uses the score from the highest compute setting.

• Update 2025-08-12 (PST) (AI summary of creator comment): - Will resolve to the GPT-5 high GPQA score; GPT-5 Pro is excluded.