HealthBench
healthbench · v1 · qa · judge GPT-5.4 (2026-03-05, judge)
Realistic health conversations graded against physician-written rubrics. Variants: full (5,000), hard (1,000) and consensus (3,671), reported separately. Official openai/simple-evals code with its GPT-5.4 (low reasoning) grader option.
New run
Sign in to see estimates and start runs.