Harvey Legal Agent Benchmark leaderboard
As of November 8, 2026, Kimi K3.1 Pro (Moonshot) leads Harvey Legal Agent Benchmark at 23.7%, ahead of Gemini 4 Argon at 19.6% and Grok 4.7 at 19.6%. 7 systems ranked. Every score here is published by the model's own lab; each row links to the page it comes from.
What is Harvey Legal Agent Benchmark? Harvey's Legal Agent Benchmark (LAB) tests whether an AI agent can produce legal work a lawyer could review, from a short instruction and the documents of a closed client matter. Made by Harvey, May 2026.
Top models on Harvey Legal Agent Benchmark, November 2026
| # | Model | Score | Price / 1M | Context | Source |
|---|---|---|---|---|---|
| 1 | Kimi K3.1 Pro Moonshot (Kimi) | max effort | $2 / $10 | 1M | Vendor-reported Moonshot AI |
| 2 | Gemini 4 Argon Google DeepMind | as reported | $2 / $10 | 1M | Vendor-reported Google DeepMind |
| 2 | Grok 4.7 SpaceXSI | xhigh effort | $2 / $6 | 500K | Vendor-reported SpaceXSI (xAI) |
| 4 | Grok 4.6 SpaceXSI | high effort, run by Vals | $2 / $6 | 500K | Vendor-reported SpaceXSI (xAI) |
| 5 | Grok 4.5 SpaceXSI | high effort, run by Vals | $2 / $6 | 500K | Vendor-reported SpaceXSI (xAI) |
| 6 | Gemini 3.8 Flash Google DeepMind | all-pass rate | $0.75 / $3.75 | 1M | Vendor-reported Google DeepMind |
| 7 | Gemini 3.7 Flash Deprecated Google DeepMind | all-pass rate | $0.75 / $3.75 | 1M | Vendor-reported Google DeepMind |
Harvey Legal Agent Benchmark: score against price
Can you trust Harvey Legal Agent Benchmark scores?
No independent leaderboard publishes Harvey Legal Agent Benchmark results for these models yet, so every number on this page is the lab's own report of its own model. Labs choose the harness, the effort level and the subset they report, and they rarely publish failures. Read the setting under each score before comparing two rows.
- Harvey published the tasks without an official leaderboard, and third parties run their own with different task sets, so numbers from different hosts do not line up.
- Rubrics are applied by model judges, not lawyers, at evaluation time.
- Harvey builds legal AI products and also built this benchmark.
How Harvey Legal Agent Benchmark works
The task
Each task gives the agent an instruction and a matter file (contracts, templates, emails) and asks for a deliverable such as a change-of-control analysis across 24 practice areas.
The score
Expert-written rubrics split each deliverable into pass-or-fail criteria. Under all-pass grading a task counts only when every criterion passes.
Test set
2,010 tasks across 24 practice areas plus contracting, as published by the authors.
Made by
Harvey, first published May 2026.
Official sources
- github.com/harveyai/harvey-labs
- Definition read on 2026-11-08
Sources and freshness
7 published scores from 5 sources. Each was read on its source page; the oldest reading on this page is from November 8, 2026. Prices beside each score come from our model price index and move when a vendor reprices.
- Vendor-reported Google DeepMind
- Vendor-reported SpaceXSI (xAI)
- Vendor-reported Google DeepMind
- Vendor-reported Moonshot AI
- Vendor-reported SpaceXSI (xAI)
Harvey Legal Agent Benchmark FAQ
Answers derived from the 7 published scores on this page and the current price index.