Health AI Benchmarks

Labs

Who makes the models on these boards.

Lab standings

Each lab's highest-ranked model on the Health Evals Index, this site's own calculation; squares show how many of the 4 boards its score rests on. How it works

  1. Anthropic Claude Opus 5.53 of 4 boards64.0
  2. OpenAI GPT-6 Astra3 of 4 boards62.4
  3. Meta Muse Spark 1.31 of 4 boards60.9
  4. Alibaba Qwen3.8 Max1 of 4 boards60.9
  5. Baichuan Baichuan-M31 of 4 boards59.4
  6. Moonshot AI Kimi K33 of 4 boards58.5
  7. Mistral Mistral Large 41 of 4 boards58.1
  8. SpaceX AI Grok 4.62 of 4 boards57.9
  9. DeepSeek DeepSeek V4.1 Flash2 of 4 boards55.5
  10. Xiaomi MiMo V2.6 Pro1 of 4 boards55.3
  11. Google Gemini 3.8 Flash3 of 4 boards54.8
  12. Zhipu GLM 5.32 of 4 boards53.9
  13. Thinking Machines Inkling1 of 4 boards53.1
  14. MiniMax MiniMax M32 of 4 boards43.8
  15. Microsoft MAI-Thinking-12 of 4 boards43.5
  16. NVIDIA Nemotron 3.5 Lightning1 of 4 boards34.7

Every lab