# MaxxBench → FrontierPhysics

Calibration · https://benchgap.net/calibration/maxxbench/current/to/frontierphysics/current · data retrieved 2026-11-18

FrontierPhysics is estimated from MaxxBench with a Hill curve fitted on 8 models measured on both: y = 0.0000 + (1.2000 − 0.0000)·x^6.00 / (0.95495^6.00 + x^6.00), R² = 0.84, cross-validated error 4.6 pp. It is used for 1 estimate.

| Estimated model | MaxxBench | FrontierPhysics | Source |
|---|---|---|---|
| [Muse Spark 1.2](https://benchgap.net/model/muse-spark-1-2.md) | 38.8% | 0.5% | estimated ± 4.6 pp, medium confidence |

## Fitted on

| Model | MaxxBench | FrontierPhysics |
|---|---|---|
| [Claude Opus 5](https://benchgap.net/model/claude-opus-5.md) | 63.5% | 9.5% |
| [Claude Opus 5.5](https://benchgap.net/model/claude-opus-5-5.md) | 76.4% | 27.4% |
| [DeepSeek V4.1 Flash 1017](https://benchgap.net/model/deepseek-v4-1-flash.md) | 55.6% | 3.6% |
| [Gemini 3.8 Flash](https://benchgap.net/model/gemini-3-8-flash.md) | 53.4% | 0.0% |
| [GLM-5.3](https://benchgap.net/model/glm-5-3.md) | 56.7% | 4.2% |
| [GPT-5.6 Luna](https://benchgap.net/model/gpt-5-6-luna.md) | 46.1% | 3.6% |
| [GPT-5.6 Sol](https://benchgap.net/model/gpt-5-6-sol.md) | 69.1% | 7.1% |
| [GPT-6 Astra](https://benchgap.net/model/gpt-6-astra.md) | 71.9% | 23.2% |
