LegalBench
Legal reasoning tasks written by lawyers. Run by Vals AI.
- Published by
- Vals AI
- Category
- Knowledge
- Index weight
- 0.5
- Models
- 136
- Data as of
- 9 Sept 2026
Vals AI (vals.ai).
- 1Claude Fable 588.6%
- 2Claude Fable 5.188.5%
- 3Gemini 3.1 Pro87.4%
- 4Gemini 3.7 Flash87.3%
- 5Gemini 3 Pro87.0%
- 6Gemini 3.8 Flash87.0%
- 7Claude Opus 587.0%
- 8GPT-5.6 Sol87.0%
- 9Gemini 3 Flash86.9%
- 10Gemini 3.6 Flash86.7%
- 11GPT-5.586.5%
- 12Grok 4.686.3%
- 13GPT-5.486.0%
- 14Kimi K386.0%
- 15GPT-586.0%
136 of 136
| # | |||||
|---|---|---|---|---|---|
| 1 |  Claude Fable 5Anthropic | 88.6% | – | 70.4 | – |
| 2 | Claude Fable 5.1Anthropic | 88.5% | – | 70.0 | – |
| 3 |  Gemini 3.1 ProGoogle | 87.4% | – | 60.7 | – |
| 4 | Gemini 3.7 FlashGoogle | 87.3% | – | 60.8 | – |
| 5 | Gemini 3 ProGoogle | 87.0% | – | 57.0 | – |
| 6 | Gemini 3.8 FlashGoogle | 87.0% | – | 60.1 | – |
| 7 | Claude Opus 5Anthropic | 87.0% | – | 70.0 | – |
| 8 |  GPT-5.6 SolOpenAI | 87.0% | – | 65.5 | – |
| 9 | Gemini 3 FlashGoogle | 86.9% | – | 58.5 | – |
| 10 | Gemini 3.6 FlashGoogle | 86.7% | – | 58.6 | – |
| 11 | GPT-5.5OpenAI | 86.5% | – | 63.7 | – |
| 12 |  Grok 4.6xAI | 86.3% | – | 58.7 | – |
| 13 | GPT-5.4OpenAI | 86.0% | – | 61.3 | – |
| 14 |  Kimi K3Moonshot AIopen | 86.0% | kimi/kimi-k3 | 66.5 | – |
| 15 | GPT-5OpenAI | 86.0% | – | 54.5 | – |
| 16 | Grok 4.5xAI | 86.0% | – | 54.2 | – |
| 17 | GPT-5.1OpenAI | 85.7% | – | 53.8 | – |
| 18 |  MiniMax-M3MiniMaxopen | 85.4% | minimax/MiniMax-M3 | 56.9 | – |
| 19 | Claude Opus 4.6Anthropic | 85.3% | – | 62.3 | – |
| 20 |  Muse Spark 1.2Meta | 85.3% | – | 61.3 | – |
| 21 | Claude Opus 4.7Anthropic | 85.3% | – | 65.5 | – |
| 22 | GPT-5.6 TerraOpenAI | 85.1% | – | 58.0 | – |
| 23 |  Qwen3.5 PlusAlibaba | 85.1% | – | 57.2 | – |
| 24 | Muse Spark 1.1Meta | 85.0% | – | 56.4 | – |
| 25 | Qwen3 7Alibaba | 84.9% | – | 58.9 | – |
| 26 |  GLM-5.3Zhipu AIopen | 84.8% | – | 59.9 | – |
| 27 | Kimi K2.6Moonshot AIopen | 84.7% | kimi/kimi-k2.6 | 60.6 | – |
| 28 | Claude Opus 4.5Anthropic | 84.6% | – | 60.6 | – |
| 29 | Grok 4.3xAI | 84.5% | grok/grok-4.3 | 57.8 | – |
| 30 | GLM-5.1Zhipu AIopen | 84.4% | zai/glm-5.1 | 60.1 | – |
| 31 | Gemini 2.5 ProGoogle | 84.3% | – | 54.1 | – |
| 32 | Qwen3.5 FlashAlibaba | 84.3% | – | 51.8 | – |
| 33 | Qwen3.6 PlusAlibabaopen | 84.2% | – | 59.0 | – |
| 34 | Muse SparkMeta | 84.2% | – | 65.4 | – |
| 35 | Claude Sonnet 4.5Anthropic | 84.1% | – | 52.8 | – |
| 36 | GLM-5.2Zhipu AIopen | 84.1% | zai/glm-5.2 | 57.4 | – |
| 37 | Gemini 3.5 Flash LiteGoogle | 84.1% | – | 44.0 | – |
| 38 | GLM-5Zhipu AIopen | 84.1% | – | 57.0 | – |
| 39 | GPT-5.6 LunaOpenAI | 84.0% | – | 56.5 | – |
| 40 | MiniMax-M2.7MiniMaxopen | 84.0% | minimax/MiniMax-M2.7 | 56.5 | – |
| 41 | GLM-5.3-FlashZhipu AIopen | 83.9% | – | 55.6 | – |
| 42 | Claude Sonnet 5Anthropic | 83.9% | – | 61.2 | – |
| 43 | Gemini 2.5 FlashGoogle | 83.8% | – | 48.8 | – |
| 44 | o3OpenAI | 83.8% | – | 52.9 | – |
| 45 | Gemini 3.1 Flash LiteGoogle | 83.8% | – | 49.1 | – |
| 46 | Qwen3 8Alibaba | 83.6% | – | 63.9 | – |
| 47 | Gemini 3.5 FlashGoogle | 83.6% | – | 60.4 | – |
| 48 | Claude Opus 4.8Anthropic | 83.6% | – | 64.7 | – |
| 49 | Claude Opus 4.1Anthropic | 83.5% | – | 52.6 | – |
| 50 | GLM-4.7Zhipu AIopen | 83.4% | zai/glm-4.7 | 54.9 | – |
| 51 | Grok 4xAI | 83.2% | 0709 | 57.2 | – |
| 52 | Grok 3 Mini FastxAI | 83.1% | – | 53.5 | – |
| 53 | GPT-4.1OpenAI | 83.1% | – | 46.2 | – |
| 54 | Claude Opus 4Anthropic | 83.1% | – | 50.7 | – |
| 55 | Inkling-SmallThinking Machinesopen | 83.0% | – | 55.0 | – |
| 56 | Claude Sonnet 4Anthropic | 83.0% | – | 49.5 | – |
| 57 | Claude Opus 4.5Anthropic | 82.8% | – | 58.0 | – |
| 58 | GPT-5.2OpenAI | 82.8% | – | 57.3 | – |
| 59 | Gemini 2.5 Flash 09Google | 82.6% | thinking, preview, 2025 | 52.5 | – |
| 60 | Grok 3xAI | 82.6% | grok/grok-3 | 50.0 | – |