Calibration

AA ITBench → MCP Atlas

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

MCP Atlas is estimated from AA ITBench with a Michaelis–Menten + offset curve fitted on 5 models measured on both: y = 0.0471 + 1.5461·x / (0.50271 + x), R² = 0.64, cross-validated error 3.5 pp. It is used for 2 estimates.

Estimated model	AA ITBench	MCP Atlas	Source
GPT-5.6 Sol	56.2%	86.3%	estimated ± 3.5 pp, low confidence
GPT-5.6 Terra	51.0%	82.6%	estimated ± 3.5 pp, medium confidence