Harvey's Legal Agent Benchmark
Professional Work · 2026-05-06
Harvey's Legal Agent Benchmark held-out evaluation tests whether agents can complete realistic legal work using documents, spreadsheets, presentations, and file-system tools. Scores report the percentage of tasks completed successfully.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 23.6 |
| Muse Spark 1.1 | 20.0 |
| Claude Fable 5 | 13.3 |
| Grok 4.5 | 12.9 |
| Kimi K3 | 10.8 |
| Opus 4.8 | 10.4 |
| GLM-5.2 | 7.1 |
| Sonnet 5 | 5.8 |
| Sonnet 4.6 | 5.0 |
| MiniMax M3 | 4.2 |
| DeepSeek-V4-Pro | 3.8 |
| GPT-5.5 | 3.8 |
| Gemini 3.6 Flash | 3.3 |
| Gemini 3.5 Flash | 2.5 |
| GPT-5.6 Sol | 2.5 |
| Inkling | 2.1 |
| Kimi K2.6 | 1.7 |
| Qwen3.7-Max | 1.7 |
| GPT-5.6 Luna | 1.3 |
| GPT-5.6 Terra | 0.4 |
| Grok 4.3 | 0.4 |
| Gemini 3.5 Flash-Lite | 0.0 |
| GLM-5.1 | 0.0 |
| GPT-5.4 | 0.0 |
| Laguna M.1 | 0.0 |
| Laguna XS.2 | 0.0 |
| Qwen3.7-Plus | 0.0 |
| Gemini 3.1 Pro Preview | 0.0 |