Legal Research Bench Criminal
Professional Work · 2026-06-23
This row reports the practice-area score for Criminal tasks in Vals AI Legal Research Bench.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.8 | 45.6 |
| Claude Opus 5 | 43.9 |
| GPT-5.6 Sol | 43.9 |
| Claude Fable 5 | 42.1 |
| Sonnet 5 | 40.4 |
| Sonnet 4.6 | 36.8 |
| GPT-5.5 | 36.8 |
| GPT-5.6 Terra | 36.8 |
| Grok 4.5 | 36.8 |
| Kimi K3 | 36.8 |
| Muse Spark 1.1 | 36.8 |
| GLM-5.1 | 29.8 |
| GPT-5.6 Luna | 29.8 |
| Gemini 3.5 Flash | 26.3 |
| MiniMax M3 | 26.3 |
| Gemini 3.6 Flash | 24.6 |
| Inkling-Small | 24.6 |
| GLM-5.2 | 22.8 |
| Inkling | 21.1 |
| Qwen3.7-Max | 21.1 |
| Gemini 3.1 Pro Preview | 19.3 |
| DeepSeek-V4-Pro | 15.8 |
| Grok 4.3 | 14.0 |
| Kimi K2.6 | 14.0 |
| Gemini 3.5 Flash-Lite | 12.3 |
| GPT-5.4 Mini | 5.3 |
| Laguna M.1 | 0.0 |
| Laguna XS.2 | 0.0 |