benchgap
Calibration

BrowseComp → APEX-Agents-AA

APEX-Agents-AA is estimated from BrowseComp with a Hill curve fitted on 11 models measured on both: y = 0.0575 + (1.2000 − 0.0575)·x^6.00 / (1.02956^6.00 + x^6.00), R² = 0.87, cross-validated error 4.6 pp. It is used for 24 estimates.

Estimated modelBrowseCompAPEX-Agents-AASource
Agents-A175.5%21.1%estimated ± 4.6 pp, high confidence
Agents-A1-4B66.8%13.7%estimated ± 4.6 pp, high confidence
Claude Mythos 588.0%37.8%estimated ± 4.6 pp, high confidence
DeepSeek V4 Flash 073173.2%18.8%estimated ± 4.6 pp, high confidence
dots3-note Preview83.3%30.8%estimated ± 4.6 pp, high confidence
GLM-4.752.0%7.6%estimated ± 4.6 pp, medium confidence
GLM-5.168.0%14.5%estimated ± 4.6 pp, high confidence
GPT-5.4 Pro89.3%39.9%estimated ± 4.6 pp, high confidence
GPT-5.5 Pro90.1%41.2%estimated ± 4.6 pp, high confidence
Inkling77.1%22.9%estimated ± 4.6 pp, high confidence
Inkling-Small77.4%23.2%estimated ± 4.6 pp, high confidence
Ling 3.0 Flash72.2%17.9%estimated ± 4.6 pp, high confidence
LongCat-Flash-Lite-Sparse48.6%7.0%estimated ± 4.6 pp, medium confidence
Nemotron 3.5 Lightning 30B A3B NVFP436.8%6.0%estimated ± 4.6 pp, medium confidence
Nemotron 3 Ultra44.4%6.5%estimated ± 4.6 pp, medium confidence
Ornith-1.5-35B-A3B67.6%14.2%estimated ± 4.6 pp, high confidence
Ornith-1.5-397B86.6%35.6%estimated ± 4.6 pp, high confidence
Ornith-1.5-9B56.4%8.8%estimated ± 4.6 pp, medium confidence
Qwen3.5-122B-A10B63.8%11.9%estimated ± 4.6 pp, high confidence
Qwen3.5-27B61.0%10.5%estimated ± 4.6 pp, high confidence
Qwen3.5-35B-A3B61.0%10.5%estimated ± 4.6 pp, high confidence
Qwen3.5 397B62.0%10.9%estimated ± 4.6 pp, high confidence
Beam77.4%23.2%estimated ± 4.6 pp, high confidence
Solar Pro 449.2%7.1%estimated ± 4.6 pp, medium confidence