PhysCodeBench - Physical Assertion Pass Rate: leaderboard
PhysCodeBench (2026) tests whether language models write Genesis physics simulation code for natural-language descriptions of scenes with rigid bodies, soft bodies, fluids and mechanisms. Share of expert-written physical assertions, read from the engine state, that the simulation passes, in percent; higher is better.
Metric: Physical-assertion pass rate (%): share of the 436 expert-written executable physical assertions (one to three per test instruction, read from the engine state) that pass, counting every assertion of a program that fails to execute as failed, for Genesis physics-simulation programs generated zero-shot from natural-language scene descriptions on the 200-example PhysCodeBench test set (mean of five attempts, temperature 0.1, 4,096-token cap), with the engine documentation in context (the full 100K-token corpus for long-context models, 20K tokens of BM25-retrieved sections for 32K-context models); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet | 23.8 | #752 |
| 2 | GPT-4o | 21.6 | #751 |
| 3 | Gemini 2.0 Pro | 19.7 | #671 |
| 4 | DeepSeek R1 | 17.9 | #482 |
| 5 | DeepSeek R1 Distill Qwen 32B | 16.2 | #1834 |
| 6 | QwQ-32B | 7.4 | #907 |
| 7 | PhysCodeBench Qwen-2.5-32B (checkpoint unspecified) | 0.9 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=physcodebench-physical-assertion-pass-rate · How It Works · Data refreshed daily, snapshot 2026-11-21.