# SkillsBench → MCP Atlas

Calibration · https://benchgap.net/calibration/skills-bench/current/to/mcpatlas/current · data retrieved 2026-11-18

MCP Atlas is estimated from SkillsBench with a Michaelis–Menten curve fitted on 5 models measured on both: y = 1.3343·x / (0.39369 + x), R² = 0.75, cross-validated error 6.6 pp. It is used for 2 estimates.

| Estimated model | SkillsBench | MCP Atlas | Source |
|---|---|---|---|
| [Qwen3.8 Max](https://benchgap.net/model/qwen3-8-max.md) | 70.2% | 85.5% | estimated ± 6.6 pp, low confidence |
| [Atria Dawn Preview](https://benchgap.net/model/atria-dawn-preview.md) | 66.4% | 83.8% | estimated ± 6.6 pp, low confidence |

## Fitted on

| Model | SkillsBench | MCP Atlas |
|---|---|---|
| [Hy4 preview](https://benchgap.net/model/hy4-preview.md) | 62.9% | 83.7% |
| [Ling 3.0 Flash](https://benchgap.net/model/ling-3-0-flash.md) | 44.8% | 65.5% |
| [Ling 3.1 Flash](https://benchgap.net/model/ling-3-1-flash.md) | 68.7% | 83.0% |
| [LongCat-2.5](https://benchgap.net/model/longcat-2-5.md) | 63.3% | 83.1% |
| [Muse Glimmer 30B](https://benchgap.net/model/muse-glimmer-30b.md) | 44.3% | 75.5% |
