MedQA (Vals AI) leaderboard
- o1 leads
- 96.5%
- models measured
- 47
- lead over #2
- +0.1 pts
- median
- 92.5%
Leader on MedQA (Vals AI) · read October 22
On October 22, 2026, 51 of our catalog's models had a MedQA (Vals AI) result, and OpenAI o1 led with 96.5%, ahead of OpenAI GPT-5.1 at 96.4%.
Medical licensing exam questions. Share answered correctly, as run by Vals AI. An archived board.
Role in the index
MedQA (Vals AI) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Latency, as printed | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | o1OpenAI | 96.5% | 11 s | Not benchmarked | $60.00 | |
| 2 | GPT-5.1OpenAI | 96.4% | 12 s | 40.4 | $10.00 | |
| 3 | Gemini 3.1 Pro PreviewGoogle | 96.4% | 39 s | 32.3 | $12.00 | |
| 4 | GPT-5OpenAI | 96.3% | 36 s | 38.8 | $10.00 | |
| 5 | GPT-5.4OpenAI | 96.1% | 23 s | 64.4 | $15.00 | |
| 6 | GPT-5 MiniOpenAI | 96.1% | 16 s | 34.2 | $2.00 | |
| 7 | o3OpenAI | 96.1% | 8.5 s | Not benchmarked | $8.00 | |
| 8 | o4 MiniOpenAI | 96.0% | 6.6 s | Not benchmarked | $4.40 | |
| 9 | Claude Opus 4.5Anthropic | 95.9% | 32 s | 44.0 | $25.00 | |
| 10 | Gemini 3 Flash PreviewGoogle | 95.8% | 26 s | 43.7 | $3.00 | |
| 11 | Claude Opus 4.6Anthropic | 95.4% | 30 s | 57.2 | $25.00 | |
| 12 | Qwen3.5 Plus 2026-04-20Qwen | 95.2% | 52 s | 39.5 | $1.80 | |
| 13 | o3 MiniOpenAI | 94.8% | 8.4 s | Not benchmarked | $4.40 | |
| 14 | Claude Sonnet 4.5Anthropic | 94.7% | 36 s | 39.5 | $15.00 | |
| 15 | Grok 4.20xAI | 94.5% | 12 s | 33.3 | $2.50 | |
| 16 | Kimi K2.5Moonshot AI | 94.4% | 42 s | 38.6 | $2.85 | |
| 17 | GLM 5Z.ai | 94.3% | 1m 21s | 41.7 | $3.20 | |
| 18 | GPT-5.2OpenAI | 94.1% | 19 s | 55.9 | $14.00 | |
| 19 | DeepSeek V3.2DeepSeek | 93.9% | 54 s | 30.0 | $0.88 | |
| 20 | GLM 4.7Z.ai | 93.7% | 1m 06s | 55.1 | $2.20 | |
| 21 | Claude Opus 4.1Anthropic | 93.6% | 24 s | Not benchmarked | $75.00 | |
| 22 | GPT-5 NanoOpenAI | 93.3% | 20 s | Not benchmarked | $0.40 | |
| 23 | Kimi K2 ThinkingMoonshot AI | 92.6% | 3m 45s | Not benchmarked | $2.50 | |
| 24 | MiniMax M2.5MiniMax | 92.5% | 34 s | 37.2 | $1.20 | |
| 25 | GLM 4.6Z.ai | 92.2% | 29 s | 8.8 | $2.20 | |
| 26 | Claude Sonnet 4.6Anthropic | 92.1% | 35 s | 57.3 | $15.00 | |
| 27 | gpt-oss-120bOpenAI | 91.4% | 12 s | Not benchmarked | $0.50 | |
| 28 | GPT-4.1OpenAI | 91.2% | 3.1 s | Not benchmarked | $8.00 | |
| 29 | MiniMax M2.1MiniMax | 91.2% | 36 s | Not benchmarked | $1.20 | |
| 30 | R1RetiredDeepSeek | 90.8% | 42 s | Not benchmarked | $2.50 | |
| 31 | GLM 4.5Z.ai | 90.0% | 60 s | Not benchmarked | $2.20 | |
| 32 | Gemini 2.5 Flash LiteGoogle | 88.9% | 15 s | Not benchmarked | $0.40 | |
| 33 | GPT-4o (2024-08-06)OpenAI | 88.2% | 3.4 s | Not benchmarked | $10.00 | |
| 34 | GPT-4.1 MiniOpenAI | 84.6% | 1.9 s | Not benchmarked | $1.60 | |
| 35 | Kimi K2 0711Moonshot AI | 84.0% | 11 s | Not benchmarked | $2.30 | |
| 36 | Grok 3RetiredxAI | 83.8% | 7.5 s | Not benchmarked | $15.00 | |
| 37 | gpt-oss-20bOpenAI | 82.9% | 25 s | Not benchmarked | $0.15 | |
| 38 | Mistral Large 3 2512Mistral AI | 82.2% | 17 s | Not benchmarked | $1.50 | |
| 39 | DeepSeek V3 0324DeepSeek | 82.0% | 13 s | Not benchmarked | $1.07 | |
| 40 | GPT-4 TurboOpenAI | 82.0% | 8.4 s | Not benchmarked | $30.00 | |
| 41 | Gemini 2.0 FlashRetiredGoogle | 81.5% | 1.9 s | Not benchmarked | $0.40 | |
| 42 | Command ACohere | 80.5% | 2.9 s | Not benchmarked | $10.00 | |
| 43 | Claude Haiku 4.5Anthropic | 79.6% | 24 s | 35.6 | $5.00 | |
| 44 | Mistral Medium 3.1Mistral AI | 78.2% | 7.3 s | Not benchmarked | $2.00 | |
| 45 | Gemini 1.5 ProRetiredGoogle | 76.5% | 5.9 s | Not benchmarked | $5.00 | |
| 46 | Mistral LargeMistral AI | 76.2% | 4.9 s | Not benchmarked | $6.00 | |
| 47 | GPT-4o-miniOpenAI | 72.4% | 2.3 s | Not benchmarked | $0.60 | |
| 48 | GPT-4.1 NanoOpenAI | 68.2% | 1.4 s | Not benchmarked | $0.40 | |
| 49 | GPT-3.5 TurboOpenAI | 58.5% | 1.6 s | Not benchmarked | $1.50 | |
| 50 | Llama 4 ScoutMeta Llama | 50.9% | 4.8 s | Not benchmarked | $0.50 | |
| 51 | Llama 4 MaverickMeta Llama | 43.3% | 7.8 s | Not benchmarked | $1.00 |
Only models in our catalog are listed, and each appears once with its best published run. 137 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out. Every other column is a figure the board publishes for the same run.