Benchmarks · agentic
Harvey's Legal Agent Benchmark
Tests an agent's ability to complete legal work using documents, spreadsheets, presentations, and file-system tools
22 models · 22 results · updated 2026-10-27 · vals.ai ↗
Leaderboard
Gemini 4 Argon high
More from this benchmark
1 published measuresTime per task30 results · lower is better
All results
22 rows| # | Model | Effort | Score | Subscription $ / task | API $ / task | Time |
|---|---|---|---|---|---|---|
| 1 | Gemini 4 Argon | high | 27.1% | $3.76 | $11 | 29 min |
| 2 | Grok 4.7 | xhigh | 18.3% | $1.00 | $11 | 36 min |
| 3 | Gemini 4 Carbon | high | 17.5% | $6.52 | $6.52 | 29 min |
| 4 | Mistral Large 4 | high | 17.5% | $4.18 | $4.18 | 58 min |
| 5 | GLM 5.3 | max | 14.6% | $0.41 | $3.99 | 32 min |
| 6 | DeepSeek V4.1 Flash 1017 | max | 14.2% | free | $0.61 | 22 min |
| 7 | MiMo V2.6 Pro | high | 13.8% | $0.18 | $0.22 | 31 min |
| 8 | MiMo V2.6 Flash | high | 13.3% | free | $0.086 | 22 min |
| 9 | Kimi K3 | max | 12.9% | $0.32 | $4.14 | 43 min |
| 10 | Gemini 3.8 Flash | high | 12.1% | $0.15 | $3.66 | 19 min |
| 11 | DeepSeek V4 Pro 0813 | max | 10.8% | $0.80 | $0.80 | 11 min |
| 12 | Sonnet 5.5 | max | 10.4% | $0.25 | $17 | 1.2 h |
| 13 | GLM 5.3 Flash | max | 10.4% | free | $0.32 | 12 min |
| 14 | Opus 5.5 | max | 9.2% | $0.32 | $21 | 58 min |
| 15 | GPT-6 Astra | max | 7.1% | $2.18 | $25 | 33 min |
| 16 | GPT-6.1 Sol | max | 6.3% | $0.41 | $3.54 | 40 min |
| 17 | Mercury 3 | high | 5.0% | $0.19 | $0.19 | 15 min |
| 18 | Haiku 5.5 | max | 4.2% | $0.090 | $4.26 | 44 min |
| 19 | GPT-6 Luna | max | 4.2% | $0.044 | $0.30 | 22 min |
| 20 | Inkling | max | 3.8% | free | $1.25 | 8.8 min |
| 21 | Inkling Small | max | 2.9% | free | $0.18 | 7.6 min |
| 22 | GPT-6.1 Terra | max | 1.7% | $1.48 | $1.48 | 14 min |