Legal Research Bench
Professional Work · 2026-06-23
Vals AI Legal Research Bench evaluates legal research performance on private tasks spanning multiple practice areas.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Carbon | 55.8 |
| Claude Opus 5 | 55.3 |
| Muse Spark 1.3 | 55.3 |
| Gemini 4 Argon | 54.8 |
| Muse Wonder Full | 53.4 |
| Muse Wonder | 50.5 |
| GLM-5.3 | 49.0 |
| GPT-5.6 Sol | 48.1 |
| Grok 4.6 | 48.1 |
| Qwen3.8-Max | 47.6 |
| Grok 4.7 | 47.1 |
| MiMo-V2.6-Pro | 47.1 |
| Kimi K3 | 46.2 |
| Qwen3.8 Max (0902) | 46.2 |
| DeepSeek V4.1 Flash 1017 | 45.2 |
| GLM-5.3 Flash | 45.2 |
| Hy4 Preview | 45.2 |
| Ember-1 | 44.7 |
| Claude Haiku 5.5 | 43.8 |
| Opus 4.8 | 43.8 |
| Muse Spark 1.2 | 43.8 |
| Sonnet 5 | 41.8 |
| DeepSeek V4.1 Flash | 41.3 |
| GPT-5.6 Terra | 41.3 |
| DeepSeek V4 Pro 0813 | 40.9 |
| GPT-5.5 | 40.4 |
| Step 5 Preview | 40.4 |
| GPT-6 Astra | 39.4 |
| Gemini 3.8 Flash | 38.9 |
| Opus 4.7 | 38.5 |
| Sonnet 4.6 | 38.5 |
| GPT-6.1 Sol | 38.5 |
| Muse Spark Open | 38.5 |
| Grok 4.5 | 38.0 |
| MiMo-V2.6-Flash | 38.0 |
| Muse Spark 1.1 | 38.0 |
| Kimi K3.1 Pro | 37.5 |
| GPT-5.6 Luna | 36.5 |
| Qwen3.8 27B | 36.1 |
| Gemini 3.7 Flash | 34.6 |