Vals AI Legal Research Bench: leaderboard
Metric: Accuracy (%). Source: www.vals.ai. 57 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 | 55.29 |
| 2 | Claude Fable 5.1 (Max) | 55.29 |
| 3 | Claude Fable 5 (Max) | 49.52 |
| 4 | GLM-5.3 (Max) | 49.04 |
| 5 | GPT-5.6 Sol (Max) | 48.08 |
| 6 | Grok 4.6 (High) | 48.08 |
| 7 | Qwen 3.8 Max | 47.6 |
| 8 | GLM-5.3 Flash (Max) | 45.19 |
| 9 | Kimi K3 | 44.23 |
| 10 | Claude Opus 4.8 (Max) | 43.75 |
| 11 | Muse Spark 1.2 (xHigh) | 43.75 |
| 12 | Claude Sonnet 5 (Max) | 41.83 |
| 13 | GPT-5.6 Terra (Max) | 41.35 |
| 14 | DeepSeek V4 Pro (0813) (Max) | 40.87 |
| 15 | GPT-5.5 (xHigh) | 40.38 |
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-legal-research-bench · How It Works · Data refreshed daily, snapshot 2026-09-05.