CritPt: leaderboard
CritPt sets 71 composite physics research challenges, created by over 50 active researchers across 11 subfields and decomposed into about 190 checkpoint tasks, with machine checkable answers scored by an automated pipeline. This board mirrors the benchmarklist.com table, whose rows mix verified, imported and self-reported entries. Accuracy in percent, higher is better. Source: arxiv.org/abs/2509.26574.
Metric: Accuracy (imported). Source: artificialanalysis.ai. Source: benchmarklist.com. Saturation forecast: Around October 2027. 279 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Muse Wonder Full (Max) | 32.9 | #10 (Muse Wonder Full) |
| 2 | GPT-5.6 Sol (Max) | 32.3 | #19 (GPT-5.6 Sol) |
| 3 | GPT-6 Astra (Max) | 31.7 | #6 (GPT-6 Astra) |
| 4 | GPT-6.1 Sol (Max) | 31.7 | #7 (GPT-6.1 Sol) |
| 5 | GPT-6.1 Sol (xHigh) | 31.7 | #7 (GPT-6.1 Sol) |
| 6 | GPT-6 Astra (xHigh) | 31.4 | #6 (GPT-6 Astra) |
| 7 | GPT-5.5 Pro (xHigh) | 30.6 | #17 (GPT-5.5 Pro) |
| 8 | GPT-5.4 Pro (xHigh) | 30 | #32 (GPT-5.4 Pro) |
| 9 | Claude Fable 5.1 (Max) | 29.7 | #9 (Claude Fable 5.1) |
| 10 | Claude Opus 5 (Max) | 29.1 | #12 (Claude Opus 5) |
| 11 | Muse Wonder (Max) | 28.3 | #15 (Muse Wonder) |
| 12 | GPT-5.5 | 27.1 | #31 |
| 13 | Gemini 4 Argon (High) | 27.1 | #8 (Gemini 4 Argon) |
| 14 | MiMo-V2.6-Pro | 26.6 | #42 |
| 15 | Gemini 3 Deep Think | 25.7 | #27 |
No result here: #3 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=critpt · How It Works · Data refreshed daily, snapshot 2026-10-20.