AA CritPt: leaderboard
CritPt sets 71 composite physics challenges, one of them an example, written by over 50 active researchers across 11 subfields and each modelled on an entry level research project. Final answers are numeric arrays, symbolic expressions or Python functions that an automated grader can check. Artificial Analysis runs the 70 test challenges itself, five repeats each, and reports accuracy in percent, higher is better.
Metric: Accuracy (%). Source: artificialanalysis.ai. Source: www-cdn.anthropic.com. Saturation forecast: Around August 2027. 583 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Muse Wonder Full (Max) | 32.86 | #10 (Muse Wonder Full) |
| 2 | GPT-5.6 Sol (Max) | 32.29 | #19 (GPT-5.6 Sol) |
| 3 | GPT-6 Astra (Max) | 31.71 | #6 (GPT-6 Astra) |
| 4 | GPT-6.1 Sol (Max) | 31.71 | #7 (GPT-6.1 Sol) |
| 5 | GPT-6.1 Sol (xHigh) | 31.71 | #7 (GPT-6.1 Sol) |
| 6 | GPT-6 Astra (xHigh) | 31.43 | #6 (GPT-6 Astra) |
| 7 | GPT-6 Sol (Max) | 30.86 | #14 (GPT-6 Sol) |
| 8 | GPT-5.5 Pro (xHigh) | 30.57 | #17 (GPT-5.5 Pro) |
| 9 | GPT-5.6 Terra (Max) | 30 | #41 (GPT-5.6 Terra) |
| 10 | GPT-6.1 Sol (High) | 30 | #7 (GPT-6.1 Sol) |
| 11 | GPT-5.4 Pro (xHigh) | 30 | #32 (GPT-5.4 Pro) |
| 12 | Claude Opus 5 (Max) | 29.14 | #12 (Claude Opus 5) |
| 13 | GPT-6 Astra (Medium) | 29.14 | #6 (GPT-6 Astra) |
| 14 | GPT-6 Astra (High) | 28.86 | #6 (GPT-6 Astra) |
| 15 | Claude Mythos 5 | 28.6 | #2 |
No result here: #3 Claude Opus 5.5, #9 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=aa-critpt · How It Works · Data refreshed daily, snapshot 2026-10-20.