Clinical Benchmarks

Agentic and workflow benchmarks

5 tracked · updated August 16, 2026

Here models act instead of answering: multi-step work inside EHR systems, terminals, and administrative tools, scored on whether the task actually got done. Harness and model matter together, consistency across repeated runs is often the headline finding, and scores collapse fastest of any category when tasks get long.

HealthAgentBench

Microsoft Research · 54 agentic tasks
  1. 1Anthropic logoClaude Code (Opus 5)55%
  2. 2OpenAI logoCodex (GPT-5.6-sol)45%
  3. 3OpenAI logoCodex (GPT 5.5)42%
  4. 4Microsoft/Anthropic logoCopilot (Opus 4.8)36%
  5. 5Microsoft/OpenAI logoCopilot (GPT 5.5)35%
via HealthAgentBench leaderboard · updated 2026-07full detail

CHI-Bench

actAVA · 75 operations workflows
  1. 1Humana (harness) / Anthropic (model) logoerius + claude-opus-554.7%
  2. 2Humana / Anthropic logoerius + claude-opus-4-837.3%
  3. 3Anthropic logoclaude-code + claude-opus-537.3%
  4. 4Anthropic logoclaude-code + claude-opus-4-833.3%
  5. 5Anthropic logoclaude-code + claude-opus-4-628.0%
via CHI-Bench leaderboard · updated 2026-08full detail

PhysicianBench

academic team · 100 clinical tasks
  1. 1OpenAI logoGPT-5.546.3 ± 1.2
  2. 2Anthropic logoClaude Opus 4.631.7 ± 2.3
  3. 3Anthropic logoClaude Opus 4.729.3 ± 2.5
  4. 4OpenAI logoGPT-5.427.7 ± 1.5
  5. 5Anthropic logoClaude Sonnet 4.623.0 ± 2.6

EHR-Complex

academic team · 3,915-task test set
  1. 1OpenAI logoGPT-5.4 (high reasoning)0.65
  2. 2Google logoGemini 3.1 Pro0.63
  3. 3Moonshot AI logoKimi-K2.50.62
  4. 4AQwen3.5-397B0.62
  5. 5OpenAI logoGPT-5.4 (low reasoning)0.58

HealthAdminBench

Kinetic Systems · 135 admin tasks
  1. 1Anthropic logoClaude Opus 4.6 (computer-use agent)36.3%
  2. 2OpenAI logoGPT-5.4 (computer-use agent)26.7%
  3. 3Moonshot AI logoKimi K2.515.6%
  4. 4AQwen 3.513.3%
  5. 5Google logoGemini 3.1 Pro11.9%

Which agentic and workflow benchmarks have current frontier-model results?

5 as of August 16, 2026: HealthAgentBench (Claude Code (Opus 5) leads at 55%); CHI-Bench (erius + claude-opus-5 leads at 54.7%); PhysicianBench (GPT-5.5 leads at 46.3 ± 1.2); EHR-Complex (GPT-5.4 (high reasoning) leads at 0.65); HealthAdminBench (Claude Opus 4.6 (computer-use agent) leads at 36.3%).

The other categories sit on the index: rubric-graded benchmarks, documentation and coding benchmarks, safety benchmarks, knowledge and exam benchmarks, composite indices.