SnorkelLegal
Professional Work
Pass@1 on 200 simulated mid-market law-firm tasks requiring matter-specific grounding, procedural discipline, long-context evidence and governed updates. Native effort and agent scaffold are not disclosed.
Top models (higher is better)
| Model | Score |
|---|---|
| Grok 4.6 | 40.7 |
| GLM-5.3 | 37.8 |
| Muse Spark 1.3 | 36.5 |
| Kimi K3 | 36.1 |
| Claude Fable 5.1 | 35.9 |
| Grok 4.7 | 32.8 |
| Claude Opus 5.5 | 31.4 |
| Claude Opus 5 | 26.4 |
| Gemini 3.8 Flash | 25.3 |
| GPT-6 Astra | 20.9 |