Physical Perception on PAI-Bench
68.5PAI-Bench ScoreGPT-5
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5Size=-2026.02 | 68.5 | |
| OmniFysicsSize=3B2026.02 | 57.7 | |
| Gemini-2.5-flashSize=-2026.02 | 57.6 | |
| Qwen3-VL-8B-InstructSize=8B2026.02 | 55.4 | |
| Qwen2.5-OmniSize=3B2026.02 | 50.6 | |
| SAIL-VL2Size=2B2026.02 | 48.1 | |
| Ovis2.5Size=2B2026.02 | 42.7 | |
| Claude-4.5-HaikuSize=-2026.02 | 41.7 | |
| InternVL3.5-8BSize=8B2026.02 | 40.3 |