Benchmarks · Science & research · Agents & computer use
FrontierPhysics
Uses the published verifier pass rate averaged across three trials per task. Score: verifier pass rate. Collected by Jake Boggs' benchmark dashboard.
22 models · 26 results · data as of 2026-11-15 · benchflow.ai ↗
Leaderboard
Opus 5.5 max
All results
26 rows| # | Model | Effort | FrontierPhysics |
|---|---|---|---|
| 1 | Opus 5.5 | max | 45.8% |
| 2 | GPT-6 Astra | max | 39.3% |
| 3 | GPT-6 Astra | xhigh | 38.7% |
| 4 | GPT-6.1 Sol | max | 33.3% |
| 5 | Gemini 4 Argon | high | 31.5% |
| 6 | Fable 5.1 | max | 31.0% |
| 7 | GPT-6.1 Terra | max | 28.6% |
| 8 | GPT-6 Sol | max | 27.4% |
| 9 | Opus 5 | max | 20.2% |
| 10 | GPT-5.6 Sol | max | 19.6% |
| 11 | Grok 4.7 | xhigh | 16.7% |
| 12 | GPT-5.6 Terra | max | 16.1% |
| 13 | Haiku 5.5 | max | 15.5% |
| 14 | GPT-5.6 Sol | xhigh | 15.5% |
| 15 | Kimi K3.1 | max | 15.5% |
| 16 | GPT-5.6 Luna | max | 12.5% |
| 17 | Kimi K3 | max | 11.3% |
| 18 | DeepSeek V4.1 Flash | max | 10.7% |
| 19 | Grok 4.6 | 10.7% | |
| 20 | MiMo V2.6 Pro | max | 10.1% |
| 21 | GPT-5.6 Terra | xhigh | 9.5% |
| 22 | GPT-5.6 Luna | xhigh | 8.9% |
| 23 | Gemini 3.8 Flash | high | 7.7% |
| 24 | GLM 5.3 | max | 7.1% |
| 25 | Mistral Large 4 | max | 5.4% |
| 26 | DeepSeek V4 Pro | max | 3.0% |