Harvey HLAB - Criteria Pass Rate - Employment Labor
Professional Work · 2026-05-06
The Employment Labor practice-area split of Harvey's Legal Agent Benchmark held-out evaluation criteria pass rate. This reports pooled rubric criteria passed within that split and is distinct from the all-pass task-score split.
Top models (higher is better)
| Model | Score |
|---|---|
| Muse Wonder Full | 91.8 |
| Muse Spark 1.3 | 89.9 |
| Muse Spark 1.2 | 89.5 |
| Gemini 4 Argon | 88.8 |
| Grok 4.7 | 88.0 |
| Muse Wonder | 88.0 |
| Grok 4.6 | 87.3 |
| Muse Spark 1.1 | 87.3 |
| GLM-5.3 | 86.1 |
| Qwen3.8-Max | 85.8 |
| Grok 4.5 | 85.4 |
| DeepSeek V4.1 Flash 1017 | 84.6 |
| Gemini 3.8 Flash | 84.6 |
| GPT-6 Astra | 84.3 |
| Mistral Large 4 Preview | 83.9 |
| Ember-1 | 83.5 |
| Kimi K3 | 83.5 |
| Claude Sonnet 5.5 | 83.1 |
| MiMo-V2.6-Pro | 83.1 |
| Opus 4.8 | 82.8 |
| Claude Haiku 5.5 | 82.4 |
| Opus 4.7 | 82.4 |
| Claude Opus 5 | 82.4 |
| GLM-5.3 Flash | 82.4 |
| Sonnet 5 | 82.0 |
| DeepSeek V4.1 Flash | 82.0 |
| GLM-5.2 | 81.3 |
| GPT-6.1 Sol | 81.3 |
| Step 5 Preview | 81.3 |
| Hy4 Preview | 80.9 |
| MiMo-V2.6-Flash | 80.9 |
| DeepSeek-V4-Flash-0731 | 80.5 |
| DeepSeek V4 Pro 0813 | 80.5 |
| Qwen3.8 27B | 79.4 |
| Gemini 3.7 Flash | 79.0 |
| Claude Opus 5.5 | 78.7 |
| Sonnet 4.6 | 78.7 |
| GPT-6 Luna | 77.9 |
| MiniMax M3 | 77.9 |
| Nemotron 3 Ultra 550B A55B (NVFP4) | 77.2 |