Health AI Benchmarks

Which AI handles health questions well?

Claude Opus 5.5 ranks first at 64.0. Claude Sonnet 5.5 is 2nd at 64.0.

The Health Evals Index is our own calculation from 4 of 6 health benchmarks, 53 models and 193 scores, 35% of them published by the model's maker. Method of 31 October 2026.

Health Evals Index

Out of 100. The squares under each column show which of the 4 benchmarks the score uses.

Tap a column to open that model. Ranks stay the same whichever filter you choose.

Each benchmark, model by model

The top 10 models on each of the 4 benchmarks in our index, on that benchmark's own 0 to 100 scale. Tap a column to read where the score came from.

Each bar is the model's lab colour. Each benchmark's page says who ran every score.

Every model on every benchmark

Index against price

43 models with a published list price, on a log price scale. The dashed line runs through models that nothing cheaper beats.

020406080100$0.1$0.3$1$3$10$30$100List price, USD per 1M tokens (3 parts input, 1 part output)Our index scoreClaude Opus 5.5, Anthropic: 64.0 at $8.00 per 1M tokens, 3 of 4 boardsClaude Fable 5, Anthropic: 63.4 at $20.00 per 1M tokens, 3 of 4 boardsClaude Opus 5, Anthropic: 63.1 at $10.00 per 1M tokens, 3 of 4 boardsGPT-6 Astra, OpenAI: 62.4 at $20.00 per 1M tokens, 3 of 4 boardsClaude Fable 5.1, Anthropic: 61.7 at $20.00 per 1M tokens, 3 of 4 boardsQwen3.8 Max, Alibaba: 60.9 at $2.48 per 1M tokens, 1 of 4 boardsGPT-6.1 Sol, OpenAI: 60.8 at $4.00 per 1M tokens, 3 of 4 boardsGPT-5.6 Sol, OpenAI: 59.6 at $8.00 per 1M tokens, 3 of 4 boardsKimi K3, Moonshot AI: 58.5 at $6.00 per 1M tokens, 3 of 4 boardsMistral Large 4, Mistral: 58.1 at $1.05 per 1M tokens, 1 of 4 boardsClaude Opus 4.8, Anthropic: 57.9 at $10.00 per 1M tokens, 2 of 4 boardsGrok 4.6, SpaceX AI: 57.9 at $3.00 per 1M tokens, 2 of 4 boardsMuse Spark 1.1, Meta: 57.9 at $2.00 per 1M tokens, 1 of 4 boardsGPT-5.6 Luna, OpenAI: 57.4 at $0.45 per 1M tokens, 3 of 4 boardsGPT-6.1 Terra, OpenAI: 57.4 at $2.00 per 1M tokens, 2 of 4 boardsGPT-5.6 Terra, OpenAI: 57.1 at $4.50 per 1M tokens, 2 of 4 boardsGPT-5.2, OpenAI: 57.0 at $4.81 per 1M tokens, 2 of 4 boardsClaude Sonnet 5, Anthropic: 56.8 at $4.00 per 1M tokens, 3 of 4 boardsGPT-6 Luna, OpenAI: 56.8 at $0.20 per 1M tokens, 2 of 4 boardsGrok 4.7, SpaceX AI: 56.5 at $3.00 per 1M tokens, 2 of 4 boardsGPT-6 Sol, OpenAI: 56.3 at $4.00 per 1M tokens, 2 of 4 boardsGPT-5.5, OpenAI: 55.7 at $11.25 per 1M tokens, 2 of 4 boardsDeepSeek V4.1 Flash, DeepSeek: 55.5 at $0.52 per 1M tokens, 2 of 4 boardsGPT-5, OpenAI: 55.0 at $3.44 per 1M tokens, 2 of 4 boardsClaude Opus 4.7, Anthropic: 54.9 at $10.00 per 1M tokens, 1 of 4 boardsGemini 3.8 Flash, Google: 54.8 at $1.50 per 1M tokens, 3 of 4 boardsGemini 4 Argon, Google: 54.6 at $4.00 per 1M tokens, 2 of 4 boardsGPT-5.4, OpenAI: 54.0 at $5.63 per 1M tokens, 2 of 4 boardsGLM 5.3, Zhipu: 53.9 at $2.15 per 1M tokens, 2 of 4 boardsInkling, Thinking Machines: 53.1 at $1.76 per 1M tokens, 1 of 4 boardsClaude Sonnet 4.6, Anthropic: 51.9 at $6.00 per 1M tokens, 1 of 4 boardsGrok 4.5, SpaceX AI: 51.3 at $3.00 per 1M tokens, 1 of 4 boardsGPT-5.1, OpenAI: 51.2 at $3.44 per 1M tokens, 2 of 4 boardsGPT-5.5 Instant, OpenAI: 51.1 at $11.25 per 1M tokens, 2 of 4 boardsGemini 3.6 Flash, Google: 50.5 at $1.50 per 1M tokens, 1 of 4 boardsClaude Haiku 4.5, Anthropic: 47.1 at $2.00 per 1M tokens, 1 of 4 boardsMiniMax M3, MiniMax: 43.8 at $0.52 per 1M tokens, 2 of 4 boardsMAI-Thinking-1, Microsoft: 43.5 at $3.50 per 1M tokens, 2 of 4 boardsGLM 5.2, Zhipu: 43.3 at $2.15 per 1M tokens, 1 of 4 boardsMistral Medium 3.5, Mistral: 37.3 at $3.00 per 1M tokens, 1 of 4 boardsClaude Haiku 5.5, Anthropic: 60.3 at $0.20 per 1M tokens, 2 of 4 boardsClaude Haiku 5.5 60.3Muse Spark 1.3, Meta: 60.9 at $2.00 per 1M tokens, 1 of 4 boardsMuse Spark 1.3 60.9Claude Sonnet 5.5, Anthropic: 64.0 at $4.00 per 1M tokens, 3 of 4 boardsClaude Sonnet 5.5 64.0Best for the price

How we put the index together

Each model's page shows the full arithmetic.

  1. Out of 100. Every score counts on its own benchmark's 0 to 100 scale.
  2. Easy and hard benchmarks. Some run easy and some hard, so each score is read against how the same models do on the others.
  3. Weights. Benchmarks built from people's own health conversations count the most.
  4. Few scores. A model with one or two scores is pulled toward the average of all models, 55.3, so the number of benchmarks appears next to every score.

The full method

Each benchmark on its own scale

The top 15 scores on each benchmark, on that benchmark's own scale. We keep the benchmarks apart because each one asks a different question.

Each bar is the model's lab colour. Each benchmark's page says who ran every score.

All boards and what they test

Each model's index score, and the benchmark scores underneath it

Read each row left to right. The squares are the model's scores on our index benchmarks, each out of 100 on that benchmark's own scale, and the diamond is where our index puts it. Ranks count all 53 models.

Scored onShowing 25 of 53 models
Our index score Benchmark score from the benchmark team or an independent group Benchmark score the model's maker published Range of its scores Average across all models, 55.3
  1. 1Claude Opus 5.5Anthropic · 3 of 4 benchmarks64.03/4
  2. 2Claude Sonnet 5.5Anthropic · 3 of 4 benchmarks64.03/4
  3. 3Claude Fable 5Anthropic · 3 of 4 benchmarks63.43/4
  4. 4Claude Opus 5Anthropic · 3 of 4 benchmarks63.13/4
  5. 5GPT-6 AstraOpenAI · 3 of 4 benchmarks62.43/4
  6. 6Claude Fable 5.1Anthropic · 3 of 4 benchmarks61.73/4
  7. 7Muse Spark 1.3Meta · 1 of 4 benchmarks60.91/4
  8. 8Qwen3.8 MaxAlibaba · 1 of 4 benchmarks60.91/4
  9. 9GPT-6.1 SolOpenAI · 3 of 4 benchmarks60.83/4
  10. 10Claude Haiku 5.5Anthropic · 2 of 4 benchmarks60.32/4
  11. 11GPT-5.6 SolOpenAI · 3 of 4 benchmarks59.63/4
  12. 12Baichuan-M3Baichuan · 1 of 4 benchmarks59.41/4
  13. 13Muse Wonder FullMeta · 1 of 4 benchmarks59.11/4
  14. 14Kimi K3Moonshot AI · 3 of 4 benchmarks58.53/4
  15. 15Muse WonderMeta · 1 of 4 benchmarks58.41/4
  16. 16Mistral Large 4Mistral · 1 of 4 benchmarks58.11/4
  17. 17Claude Opus 4.8Anthropic · 2 of 4 benchmarks57.92/4
  18. 18Grok 4.6SpaceX AI · 2 of 4 benchmarks57.92/4
  19. 19Muse Spark 1.1Meta · 1 of 4 benchmarks57.91/4
  20. 20Muse SparkMeta · 2 of 4 benchmarks57.52/4
  21. 21GPT-5.6 LunaOpenAI · 3 of 4 benchmarks57.43/4
  22. 22GPT-6.1 TerraOpenAI · 2 of 4 benchmarks57.42/4
  23. 23GPT-5.6 TerraOpenAI · 2 of 4 benchmarks57.12/4
  24. 24GPT-5.2OpenAI · 2 of 4 benchmarks57.02/4
  25. 25Claude Sonnet 5Anthropic · 3 of 4 benchmarks56.83/4

Method of 31 October 2026. Tap any square to read the sentence the score came from.

How it works Download the index

How a published score reaches this page

Every number below is a count of rows in our dataset.

29

documents we read, each kept with the exact sentence and page

18system card
4model card
3official leaderboard
2launch post
2other kinds
193

published scores taken from them

117from the team that maintains the benchmark
9from researchers with no stake in the model
67from the company that makes the model
6

health benchmarks we follow

4count toward our index
53

models in our index

12scored on 3 or more benchmarks
20scored on 2
21scored on just 1

We only add a score when we can point to the document, the sentence and the page. Scores a model's maker published stay in, and they are marked so you can tell them apart.

Read the sources