Dr. Jin Zhang

FDABench: Evaluating Regulatory Judgment

The standard for evaluating realistic FDA regulatory judgement.

That’s it.

MODEL PERFORMANCE

Opus 5.546%
GPT 6 Astra35%
Grok 4.829%
Gemini 4 Argon20%
Grok 4.619%
Kimi K318%
Fable 5.1 [xhigh]31.77%
Fable 5.1 [high]28.41%
Fable 5.1 [medium]30.5%
Opus 5.5 [xhigh]32.66%
Opus 5.5 [high]33.67%
Opus 5.5 [medium]32.6%
Sonnet 5.5 [xhigh]29.81%
Sonnet 5.5 [high]27.51%
Sonnet 5.5 [medium]23.29%