benchgap
Calibration

CritPt → BioBasicsBench

BioBasicsBench is estimated from CritPt with a offset logistic curve fitted on 7 models measured on both: y = 0.3393 + (0.5840 − 0.3393) / (1 + exp(−200.00·(x − 0.2841))), R² = 0.78, cross-validated error 10.7 pp. It is used for 38 estimates.

Estimated modelCritPtBioBasicsBenchSource
Sage-A1.520.0%33.9%estimated ± 10.7 pp, low confidence
Beam 1.0 501B-A23B FP817.4%33.9%estimated ± 10.7 pp, low confidence
Beam 1.0 501B-A23B NVFP415.4%33.9%estimated ± 10.7 pp, low confidence
Celeris-10.0%33.9%estimated ± 10.7 pp, low confidence
Exaone 4.0 1.2B0.0%33.9%estimated ± 10.7 pp, low confidence
Exaone 4.0 32B0.0%33.9%estimated ± 10.7 pp, low confidence
Gemma 4 12B0.0%33.9%estimated ± 10.7 pp, low confidence
Gemma 4 26B A4B0.0%33.9%estimated ± 10.7 pp, low confidence
Gemma 4 E2B0.0%33.9%estimated ± 10.7 pp, low confidence
GPT-OSS 20B1.4%33.9%estimated ± 10.7 pp, low confidence
Granite-4.0-350M0.0%33.9%estimated ± 10.7 pp, low confidence
Granite-4.0-H-1B0.0%33.9%estimated ± 10.7 pp, low confidence
Granite-4.0-H-350M0.0%33.9%estimated ± 10.7 pp, low confidence
Granite 4.2 30B0.3%33.9%estimated ± 10.7 pp, low confidence
LFM2.5-2.6B0.0%33.9%estimated ± 10.7 pp, low confidence
LFM2.5-8B-A1B0.0%33.9%estimated ± 10.7 pp, low confidence
LFM2.5-VL-1.6B-Extract0.0%33.9%estimated ± 10.7 pp, low confidence
Llama 3.1 405B0.0%33.9%estimated ± 10.7 pp, low confidence
Llama 3.1 70B0.0%33.9%estimated ± 10.7 pp, low confidence
Llama 3.3 70B0.0%33.9%estimated ± 10.7 pp, low confidence
Llama 4 Maverick0.0%33.9%estimated ± 10.7 pp, low confidence
Llama 4 Scout0.0%33.9%estimated ± 10.7 pp, low confidence
MiniCPM5-2B0.3%33.9%estimated ± 10.7 pp, low confidence
Nemotron 3 Nano 30B0.0%33.9%estimated ± 10.7 pp, low confidence
Nemotron 3 Nano Omni 30B A3B0.0%33.9%estimated ± 10.7 pp, low confidence
Nemotron Ultra 253B0.0%33.9%estimated ± 10.7 pp, low confidence
o31.1%33.9%estimated ± 10.7 pp, low confidence
Phi-40.0%33.9%estimated ± 10.7 pp, low confidence
Phi-4-reasoning0.0%33.9%estimated ± 10.7 pp, low confidence
Quasar 438B9.4%33.9%estimated ± 10.7 pp, low confidence
Qwen3.5-4B0.0%33.9%estimated ± 10.7 pp, low confidence
Qwen3-Omni-30B-A3B-Instruct0.0%33.9%estimated ± 10.7 pp, low confidence
Qwen3-Omni-30B-A3B-Thinking0.0%33.9%estimated ± 10.7 pp, low confidence
Sarvam 105B0.0%33.9%estimated ± 10.7 pp, low confidence
Sarvam 30B0.3%33.9%estimated ± 10.7 pp, low confidence
Trinity-Large-Preview0.9%33.9%estimated ± 10.7 pp, low confidence
Trinity-Large-Thinking0.9%33.9%estimated ± 10.7 pp, low confidence
Ultravox v0.6 Llama 3.3 70B0.0%33.9%estimated ± 10.7 pp, low confidence

Fitted on

ModelCritPtBioBasicsBench
GPT-5.211.6%31.2%
GPT-5.423.4%33.2%
GPT-5.527.1%35.4%
GPT-5.6 Sol32.3%45.6%
GPT-6.1 Sol31.7%62.6%
GPT-6 Astra31.7%67.0%
GPT-6 Luna19.4%37.6%