BenchLeader
Menu
LeaderboardsModelsProvidersBenchmarksCompareReleases8BlogMethodology

Harvey LAB

Legal work drawn from practice, built with Harvey. Run by Artificial Analysis.

As of 22 Sept 2026, Muse Spark 1.3 leads Harvey LAB on BenchLeader with 95.5%, ahead of Kimi K3 at 94.6%, across 21 model configurations with a published result.

Published by
Harveydata via Artificial Analysis
Category
Agents & tools
Index weight
Reference only
Models
21
Data as of
22 Sept 2026

Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.

Top 15
  1. 1Muse Spark 1.3 (xhigh)95.5%
  2. 2Kimi K3 (max)94.6%
  3. 3Qwen3.8 Max (0902) (max)93.7%
  4. 4Claude Fable 5 (thinking)93.6%
  5. 5Claude Opus 5 (max)93.5%
  6. 6Claude Fable 5.1 (xhigh)93.3%
  7. 7Claude Fable 5.1 (thinking)93.0%
  8. 8Claude Fable 5.1 (high)93.0%
  9. 9Claude Fable 5.1 (medium)92.7%
  10. 10Claude Opus 5.5 (xhigh)91.2%
  11. 11Claude Opus 5.5 (thinking)91.2%
  12. 12Claude Opus 5.5 (high)90.9%
  13. 13Claude Sonnet 5 (max)90.1%
  14. 14MiniMax M388.4%
  15. 15GPT-5.6 Luna (max)87.9%
21 of 21
#
1Muse Spark 1.3xhighMeta95.5%67.92026-09-02
2Kimi K3maxMoonshot AIopen ↗94.6%66.72026-07-16
3Qwen3.8 Max (0902)maxAlibaba93.7%66.22026-09-02
4Claude Fable 5thinkingAnthropic93.6%70.32026-06-09
5Claude Opus 5maxAnthropic93.5%69.52026-07-24
6Claude Fable 5.1xhighAnthropic93.3%71.12026-09-01
7Claude Fable 5.1thinkingAnthropic93.0%71.02026-09-01
8Claude Fable 5.1highAnthropic93.0%71.02026-09-01
9Claude Fable 5.1mediumAnthropic92.7%69.12026-09-01
10NewClaude Opus 5.5xhighAnthropic91.2%70.32026-09-17
11NewClaude Opus 5.5thinkingAnthropic91.2%70.72026-09-22
12NewClaude Opus 5.5highAnthropic90.9%69.92026-09-17
13Claude Sonnet 5maxAnthropic90.1%60.02026-06-30
14MiniMax M3MiniMaxopen ↗88.4%57.22026-06-01
15GPT-5.6 LunamaxOpenAI87.9%59.92026-07-09
16GPT-5.6 SolmaxOpenAI87.2%68.52026-07-09
17GPT-5.6 TerramaxOpenAI85.2%64.82026-07-09
18Gemini 3.5 FlashhighGoogle82.1%63.42026-05-19
19Nemotron 3 Ultra 550B A55BthinkingNVIDIAopen ↗81.7%57.32026-06-04
20Mistral Medium 3.5Mistral AIopen69.1%51.12026-04-29
21gpt-oss-120bhighOpenAIopen ↗13.9%49.22025-08-05

Cite as: BenchLeader, “Harvey LAB leaderboard”, https://www.benchleader.com/benchmarks/aa_harvey_lab, data as of 22 Sept 2026.

Harvey LAB: questions

What does Harvey LAB measure?
Legal work drawn from practice, built with Harvey. Run by Artificial Analysis. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads Harvey LAB?
Muse Spark 1.3 leads Harvey LAB with 95.5% as of 22 Sept 2026, ahead of Kimi K3 at 94.6%.
How many models have Harvey LAB results?
21 model configurations have a Harvey LAB result on BenchLeader, all taken from Harvey via Artificial Analysis.
Who runs Harvey LAB and how often is it updated?
Harvey LAB is published by Harvey. BenchLeader re-reads the published results every morning and records the date each result was published.
Does Harvey LAB count toward the BenchLeader Index?
No. Harvey LAB is shown for reference but left out of the composite index.