GPT-5.1-Codex-Max Evaluation Results

Gives a measured 50% time horizon of about 2h42m for a leading coding model, on trend with GPT-5. Engineers can gauge autonomous task length and the evaluator's extrapolated risk, noting OpenAI approved the post.

• articleSummary of METR's predeployment evaluation of GPT-5.6 Solmetr.org
• article前沿 AI 风险报告（2026 年 2–3 月）metr.org
• articleThe twilight of the chatbotsEthan Mollick

Checking sign-in…

Loading comments…