benchgap
Calibration

BrowseComp → AA Agentic Index

AA Agentic Index is estimated from BrowseComp with a inverse Michaelis–Menten curve fitted on 21 models measured on both: y = 0.19446·(x − 0.0000) / (0.0000 + 1.2581 − x), R² = 0.72, cross-validated error 8.2 pp. It is used for 10 estimates.

Estimated modelBrowseCompAA Agentic IndexSource
LongCat-2.591.9%52.7%estimated ± 8.2 pp, medium confidence
GPT-5.5 Pro90.1%49.1%estimated ± 8.2 pp, medium confidence
GPT-5.4 Pro89.3%47.6%estimated ± 8.2 pp, medium confidence
Sage-A1.581.7%36.0%estimated ± 8.2 pp, medium confidence
Ling 3.0 Flash Sante80.2%34.2%estimated ± 8.2 pp, medium confidence
Agents-A175.5%29.2%estimated ± 8.2 pp, medium confidence
Agents-A1-4B66.8%22.0%estimated ± 8.2 pp, medium confidence
Sarvam 105B49.5%12.6%estimated ± 8.2 pp, medium confidence
LongCat-Flash-Lite-Sparse48.6%12.2%estimated ± 8.2 pp, medium confidence
Sarvam 30B35.5%7.6%estimated ± 8.2 pp, low confidence

Fitted on

ModelBrowseCompAA Agentic Index
Claude Opus 4.779.3%39.5%
Claude Opus 4.884.3%42.6%
Claude Opus 590.8%56.2%
Claude Sonnet 584.7%44.3%
DeepSeek V4 Flash 073173.2%41.7%
DeepSeek V4 Pro83.4%49.6%
GLM-5.168.0%25.2%
GPT-5.584.4%37.3%
GPT-5.6 Luna83.3%42.7%
GPT-5.6 Sol92.2%50.5%
GPT-5.6 Terra87.5%43.7%
GPT-6 Astra91.5%51.5%
Inkling77.1%24.3%
Inkling-Small77.4%25.0%
Kimi K2.683.2%22.1%
Kimi K391.2%50.6%
Ling 3.0 Flash72.2%21.0%
MiniMax M383.5%30.8%
Nemotron 3.5 Lightning 30B A3B NVFP436.8%6.1%
Nemotron 3 Ultra44.4%21.7%
Qwen3.5-122B-A10B63.8%9.6%