Physical Reasoning - MVPBench: leaderboard

Physical-reasoning video slice measuring model accuracy on MVPBench minimal video pair tasks.

Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around April 2027. 7 models tracked.

Top models

#ModelScoreOverall rank
1GPT-4o32.5#343
2Gemini 1.5 Pro29.6#445

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=physical-reasoning-mvpbench · How It Works · Data refreshed daily, snapshot 2026-10-20.