# Toolathlon → OSWorld-Verified

Calibration · https://benchgap.net/calibration/1523 · data retrieved 2026-10-29

OSWorld-Verified is estimated from Toolathlon with a Michaelis–Menten curve fitted on 11 models measured on both: y = 1.3601·x / (0.44231 + x), R² = 0.57, cross-validated error 10.2 pp. It is used for 0 estimates.

## Fitted on

| Model | Toolathlon | OSWorld-Verified |
|---|---|---|
| [Claude Opus 4.5](https://benchgap.net/model/claude-opus-4-5.md) | 43.5% | 66.3% |
| [Claude Opus 4.8](https://benchgap.net/model/claude-opus-4-8.md) | 59.9% | 83.4% |
| [Gemini 3.5 Flash](https://benchgap.net/model/gemini-3-5-flash.md) | 56.5% | 78.4% |
| [GPT-5.4](https://benchgap.net/model/gpt-5-4.md) | 54.6% | 75.0% |
| [GPT-5.4 mini](https://benchgap.net/model/gpt-5-4-mini.md) | 42.9% | 72.1% |
| [GPT-5.4 nano](https://benchgap.net/model/gpt-5-4-nano.md) | 35.5% | 39.0% |
| [GPT-5.5](https://benchgap.net/model/gpt-5-5.md) | 55.6% | 78.7% |
| [Kimi K2.6](https://benchgap.net/model/kimi-2-6.md) | 50.0% | 73.1% |
| [Kimi K2.5](https://benchgap.net/model/kimi-k2-5.md) | 27.8% | 63.3% |
| [Muse Spark 1.1](https://benchgap.net/model/muse-spark-1-1.md) | 75.6% | 80.8% |
| [Qwen3.5 397B](https://benchgap.net/model/qwen3-5-397b.md) | 36.3% | 62.2% |