Benchmarks
MedHELM (arXiv) · MedHELM (Nature Medicine) · medhelm.org leaderboard
ARISE · MAST framework · MAST leaderboard · State of Clinical AI Report 2026
HealthBench (OpenAI) · HealthBench Professional (arXiv)

The gap
Knowledge–Practice Gap, 39 benchmarks (JMIR, Dec 2025)
Healthcare LLM Benchmarks Are Only as Good as Their Explicit Assumptions (CMU, 2026) · blog version
Abaluck et al. — Does LLM Assistance Improve Healthcare Delivery? (NBER w34660)
Deployment-Centered Evaluation (arXiv) · NEJM AI clinical-reasoning benchmark

Usage
OpenEvidence — 1M consultations/day (Mar 2026) · ~⅔ of US physicians (secondary)
Doximity 2026 State of AI in Medicine · press release

Companions
The six-layer monitoring timeline · The interactive lens quadrant · Essay: Model evaluation should look like product analytics