MedCode (Vals AI) leaderboard
- Claude Opus 5 leads
- 63.6%
- models measured
- 83
- lead over #2
- +4.5 pts
- median
- 43.3%
Leader on MedCode (Vals AI) · read October 21
On October 21, 2026, 83 of our catalog's models had a MedCode (Vals AI) result, and Anthropic Claude Opus 5 led with 63.6%, ahead of Google Gemini 3.1 Pro Preview at 59.1%.
Medical coding for billing: choosing ICD-10-CM diagnosis codes. Share correct, as run by Vals AI.
Role in the index
MedCode (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 | Claude Opus 5Anthropic | 63.6% | 25 s | 83.0 | $25.00 | |
| 2 | Gemini 3.1 Pro PreviewGoogle | 59.1% | 39 s | 32.3 | $12.00 | |
| 3 | Gemini 4 ArgonGoogle | 58.8% | 1m 20s | 81.0 | $10.00 | |
| 4 | Claude Fable 5Anthropic | 56.1% | 1m 31s | 79.3 | $50.00 | |
| 5 | Gemini 3 Flash PreviewGoogle | 55.9% | 44 s | 43.7 | $3.00 | |
| 6 | Gemini 3.5 FlashGoogle | 55.8% | 25 s | 39.8 | $9.00 | |
| 7 | Claude Opus 4.7Anthropic | 54.9% | 54 s | 67.5 | $25.00 | |
| 8 | Claude Fable 5.1Anthropic | 53.5% | 3m 35s | 84.2 | $50.00 | |
| 9 | Gemini 3.7 FlashGoogle | 53.4% | 9.2 s | 58.0 | $3.75 | |
| 10 | Claude Opus 4.8Anthropic | 53.2% | 1m 46s | 65.0 | $25.00 | |
| 11 | Gemini 3.6 FlashGoogle | 53.1% | 17 s | 62.5 | $3.75 | |
| 12 | Claude Sonnet 5.5Anthropic | 52.9% | 3m 59s | 87.3 | $10.00 | |
| 13 | GPT-5.1OpenAI | 52.7% | 55 s | 40.4 | $10.00 | |
| 14 | Claude Haiku 5.5Anthropic | 51.5% | 3m 46s | 70.7 | $0.50 | |
| 15 | Gemini 2.5 ProGoogle | 50.6% | 27 s | 19.1 | $10.00 | |
| 16 | Claude Opus 5.5Anthropic | 49.8% | 4m 07s | 88.8 | $20.00 | |
| 17 | GPT-5.2OpenAI | 49.8% | 2m 37s | 55.9 | $14.00 | |
| 18 | GPT-5OpenAI | 49.6% | 58 s | 38.8 | $10.00 | |
| 19 | Grok 4.7xAI | 49.5% | 3m 33s | 66.7 | $6.00 | |
| 20 | Kimi K3Moonshot AI | 49.4% | 37 s | 62.3 | $15.00 | |
| 21 | Muse Spark 1.2Meta | 49.4% | 1m 00s | 55.7 | $4.25 | |
| 22 | Claude Opus 4.5Anthropic | 49.2% | 1m 01s | 44.0 | $25.00 | |
| 23 | Claude Opus 4.6Anthropic | 49.1% | 2m 36s | 57.2 | $25.00 | |
| 24 | GPT-5.5OpenAI | 49.1% | 2m 40s | 71.3 | $30.00 | |
| 25 | GPT-6.1 SolOpenAI | 48.8% | 1m 55s | 82.4 | $10.00 | |
| 26 | GPT-6 AstraOpenAI | 48.5% | 1m 43s | 85.1 | $50.00 | |
| 27 | Gemini 3.8 FlashGoogle | 48.1% | 43 s | 61.6 | $3.75 | |
| 28 | Gemini 3.1 Flash Lite PreviewGoogle | 47.6% | 8.6 s | 22.7 | $1.50 | |
| 29 | Claude Sonnet 5Anthropic | 47.5% | 2m 15s | 57.7 | $10.00 | |
| 30 | o3OpenAI | 47.3% | 18 s | Not benchmarked | $8.00 | |
| 31 | Claude Opus 4.1Anthropic | 47.2% | 33 s | Not benchmarked | $75.00 | |
| 32 | GPT-6 SolOpenAI | 47.1% | 1m 03s | 76.0 | $10.00 | |
| 33 | MiniMax M3MiniMax | 46.3% | 1m 03s | 39.2 | $1.20 | |
| 34 | MiMo-V2.6-ProXiaomi | 45.0% | 2m 57s | 65.4 | $0.87 | |
| 35 | Grok 4.6xAI | 44.7% | 2m 07s | 65.7 | $6.00 | |
| 36 | GPT-6 LunaOpenAI | 44.7% | 1m 49s | 57.3 | $0.50 | |
| 37 | Claude Sonnet 4.5Anthropic | 44.1% | 1m 14s | 39.5 | $15.00 | |
| 38 | GPT-5.6 SolOpenAI | 44.0% | 1m 37s | 73.5 | $20.00 | |
| 39 | Gemini 3.5 Flash LiteGoogle | 43.5% | 6.0 s | 49.0 | $2.50 | |
| 40 | GPT-5.6 TerraOpenAI | 43.4% | 18 s | 63.9 | $12.00 | |
| 41 | Grok 4.5xAI | 43.3% | 60 s | 58.6 | $6.00 | |
| 42 | Hy4 previewTencent | 43.3% | 6m 39s | 44.3 | $2.2509 | |
| 43 | GPT-5 MiniOpenAI | 43.0% | 28 s | 34.2 | $2.00 | |
| 44 | GLM 5.3Z.ai | 42.9% | 2m 49s | 70.6 | $4.40 | |
| 45 | DeepSeek V4 Pro 0813DeepSeek | 42.5% | 3m 01s | 56.1 | $1.98 | |
| 46 | GPT-5.6 LunaOpenAI | 42.4% | 1m 21s | 58.0 | $1.20 | |
| 47 | GLM 5.1Z.ai | 41.6% | 1m 18s | 49.2 | $4.40 | |
| 48 | DeepSeek V4 Flash 0731DeepSeek | 41.4% | 2m 07s | 51.8 | $0.28 | |
| 49 | GPT-5.4OpenAI | 41.3% | 3m 07s | 64.4 | $15.00 | |
| 50 | T InklingThinking Machines | 41.2% | 2m 47s | 31.5 | $4.05 | |
| 51 | DeepSeek V4.1 FlashDeepSeek | 41.2% | 35 s | 62.5 | $0.78 | |
| 52 | MiMo-V2.6-FlashXiaomi | 41.1% | 1m 33s | 54.7 | $0.28 | |
| 53 | GPT-5.4 NanoOpenAI | 41.0% | 8.0 s | 44.8 | $1.25 | |
| 54 | GLM 5.2Z.ai | 40.8% | 1m 35s | 61.1 | $4.40 | |
| 55 | Mistral Large 4Mistral AI | 40.7% | 1m 09s | 56.4 | $4.18 | |
| 56 | Qwen3.8 Max (0902)Qwen | 40.7% | 6m 18s | 62.4 | $6.00 | |
| 57 | DeepSeek V4 Pro 0423DeepSeek | 40.5% | 6m 23s | 38.7 | $2.716 | |
| 58 | Gemini 2.5 FlashGoogle | 40.4% | 23 s | Not benchmarked | $2.50 | |
| 59 | Kimi K2.6Moonshot AI | 40.1% | 5m 05s | 52.1 | $4.00 | |
| 60 | Kimi K2.5Moonshot AI | 39.3% | 1m 15s | 38.6 | $2.85 | |
| 61 | Qwen3.7 MaxQwen | 38.8% | 29 s | 54.1 | $4.425 | |
| 62 | Nemotron 3 UltraNVIDIA | 38.6% | 19 s | 30.0 | $2.40 | |
| 63 | Grok 4.3xAI | 38.1% | 44 s | 37.4 | $2.50 | |
| 64 | T Inkling SmallThinking Machines | 37.9% | 3m 11s | 27.3 | $1.20 | |
| 65 | Qwen3.6 PlusQwen | 36.9% | 56 s | 44.1 | $1.95 | |
| 66 | Llama 4 MaverickMeta Llama | 36.5% | 21 s | Not benchmarked | $1.00 | |
| 67 | MiniMax M2.7MiniMax | 34.4% | 30 s | 36.2 | $1.20 | |
| 68 | Gemini 2.5 Flash LiteGoogle | 34.2% | 10 s | Not benchmarked | $0.40 | |
| 69 | MiniMax M2.1MiniMax | 34.1% | 20 s | Not benchmarked | $1.20 | |
| 70 | o4 MiniOpenAI | 33.8% | 21 s | Not benchmarked | $4.40 | |
| 71 | Mistral Medium 3.5Mistral AI | 33.8% | 35 s | 25.5 | $7.50 | |
| 72 | Qwen3.5-FlashQwen | 33.0% | 1m 03s | 41.1 | $0.26 | |
| 73 | GLM 4.7Z.ai | 32.8% | 2m 03s | 55.1 | $2.20 | |
| 74 | Claude Haiku 4.5Anthropic | 32.7% | 34 s | 35.6 | $5.00 | |
| 75 | MiMo-V2.5-ProXiaomi | 32.5% | 38 s | 31.6 | $0.9152 | |
| 76 | Ling 3.0 FlashinclusionAI | 32.3% | 9.7 s | 29.6 | $0.1215 | |
| 77 | Grok 4.20xAI | 32.2% | 17 s | 33.3 | $2.50 | |
| 78 | MiMo-V2.5Xiaomi | 31.9% | 18 s | 50.0 | $0.336 | |
| 79 | Mercury 2.5Inception | 31.3% | 7.9 s | 1.6 | $0.15 | |
| 80 | GPT-5 NanoOpenAI | 30.4% | 30 s | Not benchmarked | $0.40 | |
| 81 | Qwen3.8 27BQwen | 28.7% | 1m 29s | 74.2 | $2.55 | |
| 82 | Llama 4 ScoutMeta Llama | 23.3% | 11 s | Not benchmarked | $0.50 | |
| 83 | Command ACohere | 19.7% | 1m 44s | Not benchmarked | $10.00 |
Only models in our catalog are listed, and each appears once with its best published run. 136 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out. Every other column is a figure the board publishes for the same run.