LegalBench (Vals AI) leaderboard
- Claude Fable 5 leads
- 88.6%
- models measured
- 95
- lead over #2
- +0.1 pts
- median
- 83.5%
Leader on LegalBench (Vals AI) · read October 21
On October 21, 2026, 95 of our catalog's models had a LegalBench (Vals AI) result, and Anthropic Claude Fable 5 led with 88.6%.
Legal reasoning tasks across many kinds of legal work. Share answered correctly, as run by Vals AI.
Role in the index
We collect this benchmark and show it as supporting evidence. It is not part of the Vibe Coding Index, because it measures something other than coding and the index is about coding. Read the methodology
| # | Model | Result | Latency, as printed | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | Claude Fable 5Anthropic | 88.6% | 9.0 s | 82.1 | $50.00 | |
| 2 | Claude Fable 5.1Anthropic | 88.5% | 12 s | 89.1 | $50.00 | |
| 3 | Gemini 4 ArgonGoogle | 88.3% | 13 s | 80.5 | $10.00 | |
| 4 | Gemini 3.1 Pro PreviewGoogle | 87.4% | 10 s | 35.3 | $12.00 | |
| 5 | Gemini 3.7 FlashGoogle | 87.3% | 2.2 s | 69.9 | $3.75 | |
| 6 | Gemini 3.8 FlashGoogle | 87.0% | 3.3 s | 70.5 | $4.125 | |
| 7 | Claude Opus 5Anthropic | 87.0% | 4.4 s | 81.8 | $25.00 | |
| 8 | GPT-5.6 SolOpenAI | 87.0% | 6.2 s | 76.7 | $20.00 | |
| 9 | Gemini 3 Flash PreviewGoogle | 86.9% | 4.7 s | 75.8 | $3.00 | |
| 10 | Gemini 3.6 FlashGoogle | 86.7% | 3.0 s | 48.3 | $3.75 | |
| 11 | GPT-5.5OpenAI | 86.5% | 18 s | 67.8 | $30.00 | |
| 12 | Grok 4.6xAI | 86.3% | 27 s | 76.4 | $6.00 | |
| 13 | Kimi K3Moonshot AI | 86.2% | 7.1 s | 61.8 | $11.936 | |
| 14 | GPT-5.4OpenAI | 86.0% | 28 s | 37.7 | $15.00 | |
| 15 | GPT-5OpenAI | 86.0% | 19 s | 72.2 | $10.00 | |
| 16 | Grok 4.5xAI | 86.0% | 1m 08s | 75.5 | $6.00 | |
| 17 | GPT-5.1OpenAI | 85.7% | 7.0 s | 53.3 | $10.00 | |
| 18 | MiniMax M3MiniMax | 85.4% | 8.2 s | 57.8 | $1.20 | |
| 19 | Claude Opus 4.6Anthropic | 85.3% | 3.8 s | 66.9 | $25.00 | |
| 20 | Muse Spark 1.2Meta | 85.3% | 30 s | 65.1 | $4.25 | |
| 21 | Claude Opus 4.7Anthropic | 85.3% | 27 s | 51.1 | $25.00 | |
| 22 | GPT-5.6 TerraOpenAI | 85.1% | 2.8 s | 69.0 | $12.00 | |
| 23 | Qwen3.5 Plus 2026-04-20Qwen | 85.1% | 30 s | Not benchmarked | $1.80 | |
| 24 | Muse Spark 1.1Meta | 85.0% | 17 s | 48.9 | $4.25 | |
| 25 | Qwen3.7 MaxQwen | 84.9% | 12 s | 45.1 | $4.425 | |
| 26 | GLM 5.3Z.ai | 84.8% | 22 s | 73.1 | $1.40 | |
| 27 | Kimi K2.6Moonshot AI | 84.7% | 50 s | 44.2 | $3.41 | |
| 28 | Claude Opus 4.5Anthropic | 84.6% | 7.6 s | 66.8 | $25.00 | |
| 29 | Grok 4.3xAI | 84.5% | 5m 48s | 22.5 | $2.50 | |
| 30 | Grok 4.7xAI | 84.4% | 26 s | 75.0 | $6.00 | |
| 31 | GLM 5.1Z.ai | 84.4% | 14 s | 44.0 | $4.40 | |
| 32 | Qwen3.5-FlashQwen | 84.3% | 17 s | 6.2 | $0.26 | |
| 33 | Qwen3.6 PlusQwen | 84.2% | 19 s | 37.4 | $1.95 | |
| 34 | Claude Sonnet 4.5Anthropic | 84.1% | 13 s | 59.1 | $15.00 | |
| 35 | Gemini 3.5 Flash LiteGoogle | 84.1% | 1.7 s | 33.6 | $2.50 | |
| 36 | GLM 5.2Z.ai | 84.1% | 5.8 s | 57.7 | $7.00 | |
| 37 | GLM 5Z.ai | 84.1% | 16 s | 59.7 | $3.20 | |
| 38 | GPT-5.6 LunaOpenAI | 84.0% | 5.9 s | 61.5 | $1.20 | |
| 39 | MiniMax M2.7MiniMax | 84.0% | 8.9 s | 28.4 | $1.20 | |
| 40 | GLM 5.3 FlashZ.ai | 83.9% | 13 s | 71.0 | $0.50 | |
| 41 | Claude Sonnet 5Anthropic | 83.9% | 4.9 s | 63.5 | $10.00 | |
| 42 | Gemini 3.1 Flash Lite PreviewGoogle | 83.8% | 2.7 s | 8.1 | $1.50 | |
| 43 | o3OpenAI | 83.8% | 6.0 s | 70.7 | $8.00 | |
| 44 | Hy4 previewTencent | 83.8% | 1m 31s | 61.6 | $2.501 | |
| 45 | Qwen3.8 Max (0902)Qwen | 83.6% | 36 s | 63.5 | $6.00 | |
| 46 | Gemini 3.5 FlashGoogle | 83.6% | 35 s | 42.8 | $9.00 | |
| 47 | Claude Opus 4.8Anthropic | 83.6% | 3.1 s | 66.0 | $25.00 | |
| 48 | Claude Opus 4.1Anthropic | 83.5% | 2.8 s | 27.2 | $75.00 | |
| 49 | GLM 4.7Z.ai | 83.4% | 31 s | 33.6 | $1.98 | |
| 50 | DeepSeek V4.1 FlashDeepSeek | 83.3% | 6.5 s | 77.2 | $0.525 | |
| 51 | GPT-4.1OpenAI | 83.1% | 0.5 s | 46.5 | $8.00 | |
| 52 | Mercury 2.5Inception | 83.1% | 3.9 s | Not benchmarked | $0.15 | |
| 53 | T Inkling SmallThinking Machines | 83.0% | 18 s | 29.9 | $1.20 | |
| 54 | T InklingThinking Machines | 83.0% | 23 s | 49.7 | $4.05 | |
| 55 | GPT-5.2OpenAI | 82.8% | 17 s | 58.9 | $14.00 | |
| 56 | Qwen3.8 27BQwen | 82.4% | 48 s | 73.6 | $1.35 | |
| 57 | DeepSeek V4 Pro 0813DeepSeek | 82.4% | 20 s | 54.6 | $1.98 | |
| 58 | GPT-4o (2024-11-20)OpenAI | 82.2% | 0.5 s | 18.2 | $10.00 | |
| 59 | Claude Sonnet 4.6Anthropic | 82.1% | 22 s | 57.8 | $15.00 | |
| 60 | Nemotron 3 UltraNVIDIA | 82.1% | 5.2 s | Not benchmarked | $2.40 | |
| 61 | Gemini 2.5 Flash LiteGoogle | 82.0% | 3.1 s | Not benchmarked | $0.40 | |
| 62 | GPT-5 MiniOpenAI | 81.8% | 9.4 s | 59.8 | $2.00 | |
| 63 | Kimi K2 0711Moonshot AI | 81.5% | 1.2 s | 64.7 | $2.30 | |
| 64 | Claude Haiku 4.5Anthropic | 81.2% | 10 s | 49.8 | $5.00 | |
| 65 | GPT-4 TurboOpenAI | 80.5% | 1.4 s | Not benchmarked | $30.00 | |
| 66 | o1OpenAI | 80.4% | 8.3 s | 61.7 | $60.00 | |
| 67 | DeepSeek V4 Pro 0423DeepSeek | 80.3% | 3m 04s | 35.3 | $1.9105 | |
| 68 | Kimi K2 ThinkingMoonshot AI | 80.2% | 44 s | 63.4 | $2.50 | |
| 69 | GPT-4o (2024-08-06)OpenAI | 80.1% | 0.5 s | 23.1 | $10.00 | |
| 70 | MiniMax M2.5MiniMax | 80.0% | 16 s | 59.4 | $0.95 | |
| 71 | Command ACohere | 79.7% | 0.8 s | Not benchmarked | $10.00 | |
| 72 | Ling 3.0 FlashinclusionAI | 79.7% | 3.9 s | Not benchmarked | $0.063 | |
| 73 | GLM 4.6Z.ai | 79.6% | 30 s | 52.2 | $2.20 | |
| 74 | o4 MiniOpenAI | 79.2% | 4.0 s | 68.6 | $4.40 | |
| 75 | Mistral Large 3 2512Mistral AI | 79.1% | 1.7 s | Not benchmarked | $1.50 | |
| 76 | MiMo-V2.5Xiaomi | 78.9% | 11 s | 34.0 | $0.238 | |
| 77 | GPT-4.1 MiniOpenAI | 78.0% | 0.5 s | 28.5 | $1.60 | |
| 78 | GPT-5.4 NanoOpenAI | 77.9% | 2.7 s | Not benchmarked | $1.25 | |
| 79 | Llama 4 MaverickMeta Llama | 77.8% | 0.5 s | 18.1 | $0.6525 | |
| 80 | Grok 4.20xAI | 77.7% | 7.6 s | Not benchmarked | $2.50 | |
| 81 | DeepSeek V3 0324DeepSeek | 77.7% | 7.7 s | 42.0 | $1.14 | |
| 82 | DeepSeek V4 Flash 0731DeepSeek | 77.7% | 4.0 s | Not benchmarked | $1.32 | |
| 83 | MiMo-V2.5-ProXiaomi | 77.1% | 27 s | 40.0 | $0.609 | |
| 84 | DeepSeek V3.2DeepSeek | 76.1% | 31 s | 54.4 | $0.42 | |
| 85 | gpt-oss-120bOpenAI | 75.9% | 7.8 s | 34.5 | $0.60 | |
| 86 | GLM 4.5Z.ai | 75.6% | 16 s | 64.2 | $2.20 | |
| 87 | MiniMax M2.1MiniMax | 75.5% | 11 s | Not benchmarked | $1.20 | |
| 88 | Llama 4 ScoutMeta Llama | 72.0% | 0.4 s | 9.1 | $0.30 | |
| 89 | o3 MiniOpenAI | 71.5% | 12 s | 52.1 | $4.40 | |
| 90 | gpt-oss-20bOpenAI | 70.8% | 7.2 s | Not benchmarked | $0.09 | |
| 91 | Gemma 2 27BGoogle | 68.0% | 0.9 s | Not benchmarked | $0.65 | |
| 92 | GPT-3.5 TurboOpenAI | 64.4% | 0.5 s | Not benchmarked | $1.50 | |
| 93 | Mistral Medium 3.1Mistral AI | 62.0% | 0.7 s | Not benchmarked | $2.00 | |
| 94 | GPT-4.1 NanoOpenAI | 61.1% | 0.4 s | 8.9 | $0.40 | |
| 95 | GPT-5 NanoOpenAI | 50.1% | 23 s | 34.8 | $0.40 |
Only models in our catalog are listed, and each appears once with its best published run. 132 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.