Professional Reasoning Benchmark - Legal
Professional Work · 2025-11-13
Weighted score on legal reasoning tasks drawn from real professional practice.
Top models (higher is better)
| Model | Score |
|---|---|
| Muse Spark 1.3 | 61.6 |
| Muse Spark Open | 59.8 |
| Muse Wonder | 58.4 |
| Muse Spark 1.1 | 57.0 |
| Claude Fable 5 | 52.6 |
| Muse Spark | 52.3 |
| Opus 4.6 | 52.3 |
| Claude Fable 5.1 | 51.6 |
| GPT-5.6 Sol | 50.5 |
| GPT-5 Pro | 49.9 |
| o3-pro | 49.7 |
| Gemini 4 Argon | 49.4 |
| GPT-5.1 | 49.3 |
| GPT-5 | 49.0 |
| Gemini 3.8 Flash | 48.9 |
| o3 | 48.6 |
| GPT-6 Astra | 48.4 |
| GPT-6.1 Sol | 47.9 |
| Grok 4.7 | 47.6 |
| Claude Opus 5.5 | 47.6 |
| DeepSeek V4.1 Flash 1017 | 47.3 |
| GPT-5.2 Pro | 45.4 |
| GPT-5.4 | 44.4 |
| Opus 4.5 | 44.2 |
| Gemini 3.1 Pro Preview | 44.0 |
| Kimi K2.5 | 43.8 |
| GPT-6 Luna | 42.9 |
| Gemini 2.5 Pro | 41.4 |
| Gemini 2.5 Flash | 41.0 |
| Kimi K2 Thinking | 40.9 |
| Sonnet 4.5 | 40.8 |
| Gemini 3 Pro Preview | 40.6 |
| gpt-oss-120b | 40.2 |
| GPT-6 Sol | 38.6 |
| Qwen3 235B A22B Instruct 2507 | 38.3 |
| O4 Mini | 38.1 |
| DeepSeek-V3.1 | 37.6 |
| DeepSeek-R1-0528 | 36.6 |
| GPT-4.1 | 36.5 |
| Kimi K2 Instruct | 36.4 |