LegalBench
Professional Work · 2022-08-19
LegalBench evaluates legal reasoning across 162 open-source tasks organized into six reasoning types. Scores are reported as percentage performance across tasks.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 88.6 |
| Gemini 3.1 Pro Preview | 87.4 |
| Gemini 3 Pro Preview | 87.0 |
| Claude Opus 5 | 87.0 |
| GPT-5.6 Sol | 87.0 |
| Gemini 3 Flash Preview | 86.9 |
| Gemini 3.6 Flash | 86.7 |
| GPT-5.5 | 86.5 |
| GPT-5.4 | 86.0 |
| GPT-5 | 86.0 |
| Kimi K3 | 86.0 |
| Grok 4.5 | 86.0 |
| GPT-5.1 | 85.7 |
| MiniMax M3 | 85.4 |
| Opus 4.6 | 85.3 |
| Opus 4.7 | 85.3 |
| GPT-5.6 Terra | 85.1 |
| Qwen3.5 Plus (2026-02-15) | 85.1 |
| Muse Spark 1.1 | 85.0 |
| Qwen3.7-Max | 84.9 |
| Kimi K2.6 | 84.7 |
| Opus 4.5 | 84.6 |
| Grok 4.3 | 84.5 |
| GLM-5.1 | 84.4 |
| Gemini 2.5 Pro Experimental 03-25 | 84.3 |
| Qwen3.5-Flash | 84.3 |
| Qwen3.6 Plus (2026-04-02) | 84.2 |
| Muse Spark | 84.2 |
| Sonnet 4.5 | 84.1 |
| Gemini 3.5 Flash-Lite | 84.1 |
| GLM-5.2 | 84.1 |
| GLM-5 | 84.1 |
| GPT-5.6 Luna | 84.0 |
| MiniMax M2.7 | 84.0 |
| Sonnet 5 | 83.9 |
| Gemini 2.5 Flash Preview 04-17 | 83.8 |
| Gemini 3.1 Flash-Lite Preview | 83.8 |
| o3 | 83.8 |
| Gemini 3.5 Flash | 83.6 |
| Opus 4.8 | 83.6 |