Physical Reasoning on CosmosReason1-Bench
64.7Overall ScoreGemini-2.5-Pro
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Gemini-2.5-ProAccess=Closed-Source2026.02 | 64.7 | 45 | 84 | 83.6 | 45 | 74 | |
| PhyCritic-7BAccess=Open-Source, Parameters=7B2026.02 | 63.9 | 53 | 68 | 95.5 | 37 | 63 | |
| Cosmos-R1-7BAccess=Open-Source, Parameters=7B2026.02 | 63 | 49.4 | 63 | 83.8 | 58.8 | 60 | |
| Gemini-2.5-FlashAccess=Closed-Source2026.02 | 57.9 | 42.7 | 68.2 | 78.2 | 35.5 | 62.8 | |
| GPT-4oAccess=Closed-Source2026.02 | 56.3 | 38 | 64 | 78.2 | 33 | 66 | |
| Eagle-2.5-8BAccess=Open-Source, Parameters=8B2026.02 | 54.3 | 39 | 67 | 67.4 | 29 | 68 | |
| Qwen2.5-VL-7BAccess=Open-Source, Parameters=7B2026.02 | 54.3 | 40.4 | 50 | 82.5 | 38 | 57.6 | |
| Robobrain-2.0-7BAccess=Open-Source, Parameters=7B2026.02 | 50.4 | 37 | 41 | 85.4 | 29 | 56 |