benchgap
Calibration

SimpleQA → AA-Omniscience Accuracy

AA-Omniscience Accuracy is estimated from SimpleQA with a Michaelis–Menten curve fitted on 7 models measured on both: y = 0.4561·x / (0.19959 + x), R² = 0.86, cross-validated error 2.7 pp. It is used for 4 estimates.

Estimated modelSimpleQAAA-Omniscience AccuracySource
Grok 3 [Beta]43.6%31.3%estimated ± 2.7 pp, low confidence
DeepSeek-R1-052827.8%26.5%estimated ± 2.7 pp, medium confidence
Gemini 2.0 Flash-Lite16.5%20.6%estimated ± 2.7 pp, medium confidence
Gemini 1.5 Flash8.6%13.7%estimated ± 2.7 pp, low confidence

Fitted on

ModelSimpleQAAA-Omniscience Accuracy
DeepSeek-R130.1%30.5%
DeepSeek V324.9%25.5%
Gemini 2.0 Flash29.9%23.5%
Gemini 2.5 Flash-Lite10.7%15.4%
Kimi K231.0%27.4%
Mistral Small 3.110.4%15.1%
o4-mini20.2%24.8%