IntPhys 2
Multimodal · 2025-06-11
IntPhys 2 evaluates intuitive physics understanding with video clips that test object permanence, solidity, and spatio-temporal continuity. Models identify physically plausible versus impossible physical-world scenes.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 77.4 |
| Gemini 4 Argon | 74.6 |
| GPT-6.1 Sol | 72.3 |
| Muse Wonder | 70.2 |
| Claude Opus 5.5 | 68.3 |
| Gemini 3.7 Flash | 68.2 |
| Claude Fable 5.1 | 67.7 |
| Claude Fable 5 | 67.1 |
| Gemini 3.5 Flash | 67.0 |
| Muse Spark 1.1 | 66.8 |
| Claude Opus 5 | 65.1 |
| Muse Spark Open | 64.9 |
| GPT-5.6 Sol | 63.9 |
| Gemini 3 Flash Preview | 63.4 |
| Grok 4.20 | 62.6 |
| Gemini 3.8 Flash | 62.5 |
| Grok 4.5 | 61.4 |
| GPT-5.6 Terra | 61.3 |
| Kimi K2.6 | 60.0 |
| GPT-5.5 | 59.9 |
| GPT-5 Mini | 59.2 |
| Kimi K3 | 58.6 |
| Gemini 3.1 Flash-Lite Preview | 58.4 |
| GPT-5.2 | 58.3 |
| GPT-5.6 Luna | 58.0 |
| Gemini 2.5 Flash | 58.0 |
| Gemini 3 Pro Preview | 56.9 |
| Grok 4.3 | 56.9 |
| GPT-5.4 | 56.4 |
| Opus 4.5 | 56.3 |
| GPT-5 | 56.1 |
| Opus 4.7 | 56.0 |
| Gemini 2.5 Pro | 56.0 |
| Opus 4.8 | 55.1 |
| Grok 4.1 Fast | 55.0 |
| Grok 4 | 54.9 |
| o3 | 54.7 |
| Sonnet 4.5 | 54.4 |
| Gemini 2.5 Flash-Lite | 54.1 |
| Opus 4.6 | 53.6 |