Harvey's Legal Agent Benchmark v1
The first version of the Harvey legal agent benchmark, which Vals AI keeps online. Run by Vals AI.
As of 24 Oct 2026, Muse Spark 1.2 leads Harvey's Legal Agent Benchmark v1 on BenchLeader with 25.4%, ahead of Muse Spark 1.3 at 23.8%, across 77 model configurations with a published result.
- Published by
- Vals AI
- Category
- Agents & tools
- Index weight
- Reference only
- Models
- 77
- Data as of
- 24 Oct 2026
- Newest result
- 18 Oct 2026
Vals AI (vals.ai).
- 1Muse Spark 1.2 (xhigh)25.4%
- 2Muse Spark 1.3 (max)23.8%
- 3Muse Spark 1.3 (xhigh)22.9%
- 4Muse Wonder Full (max)21.7%
- 5Muse Spark 1.1 (xhigh)20.0%
- 6Gemini 4 Argon (high)19.6%
- 7GLM-Legal (max)19.2%
- 8Muse Wonder (max)17.1%
- 9Muse Spark Open (xhigh)16.7%
- 10Grok 4.6 (high)15.8%
- 11Mistral Large 4 (high)15.8%
- 12Kimi K3 (max)12.9%
- 13Grok 4.5 (high)12.9%
- 14Grok 4.7 (xhigh)12.5%
- 15Claude Fable 511.3%
77 of 77
| # | ||||
|---|---|---|---|---|
| 1 |  Muse Spark 1.2xhigh | 25.4% | 63.2 | 2026-10-18 |
| 2 | Muse Spark 1.3max | 23.8% | 67.1 | 2026-10-18 |
| 3 | Muse Spark 1.3xhigh | 22.9% | 64.9 | 2026-10-18 |
| 4 | Muse Wonder Fullmax | 21.7% | 70.7 | 2026-10-18 |
| 5 | Muse Spark 1.1xhigh | 20.0% | 61.6 | 2026-10-18 |
| 6 |  Gemini 4 Argonhigh | 19.6% | 70.8 | 2026-10-18 |
| 7 |  GLM-Legalmax | 19.2% | – | 2026-10-18 |
| 8 | Muse Wondermax | 17.1% | 67.5 | 2026-10-18 |
| 9 | Muse Spark Openxhighopen ↗ | 16.7% | 62.0 | 2026-10-18 |
| 10 |  Grok 4.6high | 15.8% | 62.9 | 2026-10-18 |
| 11 |  Mistral Large 4highopen | 15.8% | 49.8 | 2026-10-18 |
| 12 |  Kimi K3maxopen ↗ | 12.9% | 64.7 | 2026-10-18 |
| 13 | Grok 4.5high | 12.9% | 59.3 | 2026-10-18 |
| 14 | Grok 4.7xhigh | 12.5% | 61.6 | 2026-10-18 |
| 15 |  Claude Fable 5 | 11.3% | 67.1 | 2026-10-18 |
| 16 |  MiMo-V2.6-Flashopen ↗ | 11.3% | 59.7 | 2026-10-18 |
| 17 |  Qwen3.8 27Bxhighopen ↗ | 11.3% | 56.2 | 2026-10-18 |
| 18 | MiMo-V2.6-Proopen ↗ | 10.8% | 64.0 | 2026-10-18 |
| 19 | Qwen3.8 Maxmax | 10.4% | 63.3 | 2026-10-18 |
| 20 | Gemini 3.8 Flashhigh | 10.0% | 64.4 | 2026-10-18 |
| 21 | Claude Opus 4.8 | 9.6% | 60.1 | 2026-10-18 |
| 22 | New DeepSeek V4.1 Flash 1017maxopen ↗ | 9.2% | 64.0 | 2026-10-18 |
| 23 |  Step 5 Previewopen | 9.2% | 60.4 | 2026-10-18 |
| 24 | Ember-1 | 9.2% | – | 2026-10-18 |
| 25 | Gemini 3.7 Flashhigh | 8.8% | 63.5 | 2026-10-18 |
| 26 | GLM 5.3maxopen ↗ | 8.3% | 63.5 | 2026-10-18 |
| 27 | DeepSeek V4 Flashhighopen ↗ | 8.3% | 57.4 | 2026-10-18 |
| 28 | DeepSeek V4 Promaxopen ↗ | 7.5% | 60.8 | 2026-10-18 |
| 29 | GLM 5.2maxopen ↗ | 7.1% | 62.3 | 2026-10-18 |
| 30 | Claude Fable 5.1 | 6.7% | 66.4 | 2026-10-18 |
| 31 | Claude Opus 5 | 6.7% | 65.5 | 2026-10-18 |
| 32 | Claude Opus 4.7 | 6.7% | 63.3 | 2026-10-18 |
| 33 | DeepSeek V4.1 Flashmaxopen ↗ | 6.7% | 61.2 | 2026-10-18 |
| 34 | GLM 5.3 Flashmaxopen ↗ | 6.7% | 54.2 | 2026-10-18 |
| 35 |  GPT-6 Astramax | 5.4% | 70.4 | 2026-10-18 |
| 36 | GPT-6.1 Solmax | 5.4% | 69.9 | 2026-10-18 |
| 37 | Claude Sonnet 4.6 | 5.0% | 57.1 | 2026-10-18 |
| 38 | Claude Sonnet 5 | 5.0% | 56.1 | 2026-10-18 |
| 39 |  MiniMax M3open ↗ | 4.2% | 56.4 | 2026-10-18 |
| 40 | GPT-5.5xhigh | 3.8% | 66.1 | 2026-10-18 |
| 41 | Claude Opus 5.5 | 3.8% | 62.0 | 2026-10-18 |
| 42 | Gemini 3.6 Flashhigh | 3.3% | 62.0 | 2026-10-18 |
| 43 | GPT-6 Lunamax | 2.9% | 60.0 | 2026-10-18 |
| 44 | Claude Sonnet 5.5 | 2.9% | – | 2026-10-18 |
| 45 | GPT-5.6 Solmax | 2.5% | 68.0 | 2026-10-18 |
| 46 | Gemini 3.5 Flashhigh | 2.5% | 63.1 | 2026-10-18 |
| 47 | MiMo-V2.5-Proopen ↗ | 2.1% | 54.9 | 2026-10-18 |
| 48 | GPT-6 Solmax | 1.7% | 66.4 | 2026-10-18 |
| 49 | Qwen3.7max | 1.7% | 60.6 | 2026-10-18 |
| 50 | Kimi K2.6open ↗ | 1.7% | 55.4 | 2026-10-18 |
| 51 | Inkling Smallopen ↗ | 1.7% | 54.8 | 2026-10-18 |
| 52 | GPT-5.6 Lunamax | 1.3% | 59.9 | 2026-10-18 |
| 53 | Qwen3.6 Plus | 1.3% | 58.8 | 2026-10-18 |
| 54 |  Ling 3.0 Flashopen ↗ | 1.3% | 52.8 | 2026-10-18 |
| 55 | Claude Haiku 5.5 | 1.3% | – | 2026-10-18 |
| 56 | GPT-5.6 Terramax | 0.8% | 64.4 | 2026-10-18 |
| 57 | Claude Haiku 4.5thinking | 0.8% | 46.7 | 2026-10-18 |
| 58 | Grok 4.3high | 0.4% | 57.0 | 2026-10-18 |
| 59 |  Nemotron 3 Ultra 550B A55Bopen ↗ | 0.4% | 55.5 | 2026-10-18 |
| 60 | Mistral Medium 3.5highopen | 0.4% | – | 2026-10-18 |
More agents & tools benchmarks · All LLM benchmarks
Cite as: BenchLeader, “Harvey's Legal Agent Benchmark v1 leaderboard”, data as of 24 Oct 2026, https://www.benchleader.com/benchmarks/vals_hlab_v1.
Harvey's Legal Agent Benchmark v1: questions
- What does Harvey's Legal Agent Benchmark v1 measure?
- The first version of the Harvey legal agent benchmark, which Vals AI keeps online. Run by Vals AI. Scores are reported in percent of tasks solved; higher is better.
- Which AI model leads Harvey's Legal Agent Benchmark v1?
- Muse Spark 1.2 leads Harvey's Legal Agent Benchmark v1 with 25.4% as of 24 Oct 2026, ahead of Muse Spark 1.3 at 23.8%.
- How many models have Harvey's Legal Agent Benchmark v1 results?
- 77 model configurations have a Harvey's Legal Agent Benchmark v1 result on BenchLeader, all taken from Vals AI.
- Who runs Harvey's Legal Agent Benchmark v1 and how often is it updated?
- Harvey's Legal Agent Benchmark v1 is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does Harvey's Legal Agent Benchmark v1 count toward the BenchLeader Index?
- No. Harvey's Legal Agent Benchmark v1 is shown for reference but left out of the composite index.