Calibration

Vals MMLU-Pro → Artificial Analysis Intelligence Index

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

Artificial Analysis Intelligence Index is estimated from Vals MMLU-Pro with a linear curve fitted on 44 models measured on both: y = 2.4583·x + -1.8138, R² = 0.33, cross-validated error 10.5 pp. It is used for 0 estimates.

Fitted on

Model	Vals MMLU-Pro	Artificial Analysis Intelligence Index
Claude Fable 5	91.5%	49.6%
Claude Fable 5.1	92.4%	49.0%
Claude Opus 4.7	89.9%	30.9%
Claude Opus 4.8	89.6%	41.8%
Claude Opus 5	91.6%	50.8%
Claude Sonnet 4.6	87.3%	24.7%
Claude Sonnet 5	87.5%	38.2%
DeepSeek V4 Flash 0731	86.2%	34.3%
DeepSeek V4 Pro 0813	87.0%	53.2%
Gemini 3.5 Flash	89.5%	50.2%
Gemini 3.5 Flash-Lite	85.8%	11.7%
Gemini 3.6 Flash	89.3%	34.0%
Gemini 3.7 Flash	90.1%	39.1%
Gemini 3.8 Flash	90.2%	30.6%
Gemini 3 Flash	88.6%	17.9%
GLM-4.6	82.2%	14.9%
GLM-4.7	82.7%	22.2%
GLM-5.1	86.9%	26.1%
GLM-5.2	86.7%	33.7%
GLM-5.3	86.8%	31.7%
GPT-5.4 mini	84.6%	24.1%
GPT-5.4 nano	77.2%	20.7%
GPT-5.5	88.1%	38.4%
GPT-5.6 Luna	86.0%	51.2%
GPT-5.6 Sol	89.1%	58.9%
GPT-5.6 Terra	86.7%	54.9%
Grok 4.3	85.8%	37.6%
Grok 4.5	89.2%	38.8%
Grok 4.6	89.4%	44.3%
Inkling	86.3%	12.5%
Inkling-Small	85.6%	12.9%
Kimi K2.6	87.6%	27.0%
Kimi K3	88.0%	43.6%
Ling 3.0 Flash	82.0%	20.1%
MiMo-V2.5-Pro	84.6%	26.0%
MiniMax M2.7	80.4%	22.8%
MiniMax M3	84.2%	18.7%
Muse Spark	87.3%	31.3%
Muse Spark 1.1	88.7%	33.7%
Muse Spark 1.2	88.3%	39.6%
Nemotron 3 Ultra	85.8%	11.8%
Qwen3.6 Plus	87.7%	27.0%
Qwen3.7 Max	89.3%	29.5%
Qwen3.8-27B	84.3%	20.1%