CritPt v2 launches: reviewed physics problems, code tools, 80%+ scores
- Source
- x.com
- Date
CritPt v2 has launched on Artificial Analysis: the CritPt authors have reviewed all 70 research-level physics challenges, and models now solve them with a code-execution tool Following feedback from the research community, the CritPt team and physics expert collaborators reviewed every challenge, corrected underspecified statements, errors, and typos, and improved the stability of the grader. We now run CritPt v2 agentically: models can write and run code in an offline sandbox with tools for computation. We’ll be updating the Intelligence Index in late October, including changes to CritPt. Key takeaways: ➤ Recent OpenAI releases top the leaderboard: GPT-6 Astra (max) and GPT-6.1 Sol (max) score highest at 83% and 82%, followed by Claude Opus 5.5 (max with fallback) at 78% and GPT-5.6 Sol (max) at 75% ➤ CritPt v2 shows the rapid progress at the frontier on research-level physics: GPT-5 (high), released in August 2025, scores 29% on CritPt v2, 54 points below GPT-6 Astra (max) Thank you to @MiniHui_zhu, @MinyangTian1, @haopeng_uiuc and the broader CritPt team for their work on this revision and their continued collaboration
CritPt v2 results show frontier models above 80% on research-level physics with code tools, up from 32% on v1 for the top model. Scores are not comparable across versions, and headroom is shrinking.
- For CritPt v2, the 's authors and physics collaborators reviewed all 70 research-level challenges, fixing underspecified statements, errors and typos, and made the grader more stable after research-community feedback.
- Models now attempt CritPt v2 agentically, writing and running code in an offline that provides tools for computation.
- GPT-6 Astra and GPT-6.1 Sol lead at 83% and 82%, ahead of Claude Opus 5.5 at 78%, while GPT-5 from August 2025 scores 29%. Artificial Analysis says CritPt changes will feed its late-October Intelligence Index update.
Checking sign-in…
Loading comments…