Benchmark Almanac

Harvey LAB v1.1 (HLAB-Vals-1.1)

Gemini 4 Argon 27.1% leads since October 2026
Gemini 4 Argon 27.1% leads since October 2026
In use

Agents produce legal work product from matter files with file tools and docx, pptx and xlsx skills. A task passes only if every expert rubric criterion passes.

Harvey fixed the 120-task private holdout, removing 17 unsolvable criteria. The harness was resynced with a finish tool and a reasoning-first judge. Vals says scores are not comparable with the v1 board.

Released
October 2026
Kind
Law
Level
Expert
Scored
Graded by a model
Answers
Open-ended
Tasks
Public, plus a private test set
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 13th, 2026
38
Models scored on its own leaderboard
1
Launch post lists it, from 1 lab

Gemini 4 Argon leads with 27.1%

Best score per model from the benchmark's own leaderboard, read on November 13th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

Gemini 4 Argon leads with 27.1%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

38 models. Scroll sideways for the rest.

  1. 27.1 Gemini 4 Argon
  2. 25.0 Muse Spark Open
  3. 25.0 Muse Ember
  4. 23.8 Muse Wonder
  5. 23.3 Muse Wonder Full
  6. 21.7 Inkling 2 Small
  7. 18.3 Grok 4.7
  8. 17.5 Mistral Large 4 Preview
  9. 17.5 Gemini 4 Carbon
  10. 15.8 Mistral Large 4
  11. 15.4 Qwen3.8 Max (0902)
  12. 14.6 GLM-5.3
  13. 14.2 DeepSeek V4.1 Flash 1017
  14. 13.8 MiMo-V2.6 Pro
  15. 13.3 MiMo-V2.6 Flash
  16. 12.9 Kimi K3
  17. 12.9 Step 5 Preview
  18. 12.5 Tencent Hy4
  19. 12.1 Gemini 3.8 Flash
  20. 11.7 Kimi K3.1 Pro
  21. 11.3 Gemini 4 Hydrogen
  22. 10.8 DeepSeek V4 Pro 0813
  23. 10.4 GLM-5.3-Flash
  24. 10.4 Claude Sonnet 5.5
  25. 9.17 Claude Opus 5.5
  26. 8.75 Tencent Hy4 preview
  27. 7.50 MiniMax-M3.1-Flash
  28. 7.08 GPT-6 Astra
  29. 6.25 GPT-6.1 Sol
  30. 5.00 Mercury 3
  31. 5.00 Claude Fable 5.2
  32. 4.58 Beam 1.0
  33. 4.17 GPT-6 Luna
  34. 4.17 Claude Haiku 5.5
  35. 3.75 Inkling
  36. 3.75 Nemotron 3.5 Super VL
  37. 2.92 Inkling-Small
  38. 1.67 GPT-6.1 Terra
  1. 1 Gemini 4 Argon 27.1%
  2. 2 Muse Spark Open 25%
  3. 3 Muse Ember 25%
  4. 4 Muse Wonder 23.8%
  5. 5 Muse Wonder Full 23.3%
  6. 6 Inkling 2 Small 21.7%
  7. 7 Grok 4.7 18.3%
  8. 8 Mistral Large 4 Preview 17.5%
  9. 9 Gemini 4 Carbon 17.5%
  10. 10 Mistral Large 4 15.8%
  11. 11 Qwen3.8 Max (0902) 15.4%
  12. 12 GLM-5.3 14.6%
  13. 13 DeepSeek V4.1 Flash 1017 14.2%
  14. 14 MiMo-V2.6 Pro 13.8%
  15. 15 MiMo-V2.6 Flash 13.3%
  16. 16 Kimi K3 12.9%
  17. 17 Step 5 Preview 12.9%
  18. 18 Tencent Hy4 12.5%
  19. 19 Gemini 3.8 Flash 12.1%
  20. 20 Kimi K3.1 Pro 11.7%
  21. 21 Gemini 4 Hydrogen 11.3%
  22. 22 DeepSeek V4 Pro 0813 10.8%
  23. 23 GLM-5.3-Flash 10.4%
  24. 24 Claude Sonnet 5.5 10.4%
  25. 25 Claude Opus 5.5 9.17%
  26. 26 Tencent Hy4 preview 8.75%
  27. 27 MiniMax-M3.1-Flash 7.50%
  28. 28 GPT-6 Astra 7.08%
  29. 29 GPT-6.1 Sol 6.25%
  30. 30 Mercury 3 5.00%
  31. 31 Claude Fable 5.2 5.00%
  32. 32 Beam 1.0 4.58%
  33. 33 GPT-6 Luna 4.17%
  34. 34 Claude Haiku 5.5 4.17%
  35. 35 Inkling 3.75%
  36. 36 Nemotron 3.5 Super VL 3.75%
  37. 37 Inkling-Small 2.92%
  38. 38 GPT-6.1 Terra 1.67%

Source: Official Harvey LAB v1.1 (HLAB-Vals-1.1) leaderboard

38 models. Scroll sideways for the rest.

  1. 27.1 Gemini 4 Argon
  2. 25.0 Muse Spark Open
  3. 25.0 Muse Ember
  4. 23.8 Muse Wonder
  5. 23.3 Muse Wonder Full
  6. 21.7 Inkling 2 Small
  7. 18.3 Grok 4.7
  8. 17.5 Mistral Large 4 Preview
  9. 17.5 Gemini 4 Carbon
  10. 15.8 Mistral Large 4
  11. 15.4 Qwen3.8 Max (0902)
  12. 14.6 GLM-5.3
  13. 14.2 DeepSeek V4.1 Flash 1017
  14. 13.8 MiMo-V2.6 Pro
  15. 13.3 MiMo-V2.6 Flash
  16. 12.9 Kimi K3
  17. 12.9 Step 5 Preview
  18. 12.5 Tencent Hy4
  19. 12.1 Gemini 3.8 Flash
  20. 11.7 Kimi K3.1 Pro
  21. 11.3 Gemini 4 Hydrogen
  22. 10.8 DeepSeek V4 Pro 0813
  23. 10.4 GLM-5.3-Flash
  24. 10.4 Claude Sonnet 5.5
  25. 9.17 Claude Opus 5.5
  26. 8.75 Tencent Hy4 preview
  27. 7.50 MiniMax-M3.1-Flash
  28. 7.08 GPT-6 Astra
  29. 6.25 GPT-6.1 Sol
  30. 5.00 Mercury 3
  31. 5.00 Claude Fable 5.2
  32. 4.58 Beam 1.0
  33. 4.17 GPT-6 Luna
  34. 4.17 Claude Haiku 5.5
  35. 3.75 Inkling
  36. 3.75 Nemotron 3.5 Super VL
  37. 2.92 Inkling-Small
  38. 1.67 GPT-6.1 Terra
  1. 1 Gemini 4 Argon 27.1%
  2. 2 Muse Spark Open 25%
  3. 3 Muse Ember 25%
  4. 4 Muse Wonder 23.8%
  5. 5 Muse Wonder Full 23.3%
  6. 6 Inkling 2 Small 21.7%
  7. 7 Grok 4.7 18.3%
  8. 8 Mistral Large 4 Preview 17.5%
  9. 9 Gemini 4 Carbon 17.5%
  10. 10 Mistral Large 4 15.8%
  11. 11 Qwen3.8 Max (0902) 15.4%
  12. 12 GLM-5.3 14.6%
  13. 13 DeepSeek V4.1 Flash 1017 14.2%
  14. 14 MiMo-V2.6 Pro 13.8%
  15. 15 MiMo-V2.6 Flash 13.3%
  16. 16 Kimi K3 12.9%
  17. 17 Step 5 Preview 12.9%
  18. 18 Tencent Hy4 12.5%
  19. 19 Gemini 3.8 Flash 12.1%
  20. 20 Kimi K3.1 Pro 11.7%
  21. 21 Gemini 4 Hydrogen 11.3%
  22. 22 DeepSeek V4 Pro 0813 10.8%
  23. 23 GLM-5.3-Flash 10.4%
  24. 24 Claude Sonnet 5.5 10.4%
  25. 25 Claude Opus 5.5 9.17%
  26. 26 Tencent Hy4 preview 8.75%
  27. 27 MiniMax-M3.1-Flash 7.50%
  28. 28 GPT-6 Astra 7.08%
  29. 29 GPT-6.1 Sol 6.25%
  30. 30 Mercury 3 5.00%
  31. 31 Claude Fable 5.2 5.00%
  32. 32 Beam 1.0 4.58%
  33. 33 GPT-6 Luna 4.17%
  34. 34 Claude Haiku 5.5 4.17%
  35. 35 Inkling 3.75%
  36. 36 Nemotron 3.5 Super VL 3.75%
  37. 37 Inkling-Small 2.92%
  38. 38 GPT-6.1 Terra 1.67%

Sources: Official Harvey LAB v1.1 (HLAB-Vals-1.1) leaderboard and labs’ model cards

#ModelScoreSettingDateSource
1 Gemini 4 Argon 27.1% Sep 2026 Official leaderboard
2 Muse Spark Open 25% Oct 2026 Official leaderboard
3 Muse Ember 25% Nov 2026 Official leaderboard
4 Muse Wonder 23.8% Oct 2026 Official leaderboard
5 Muse Wonder Full 23.3% Oct 2026 Official leaderboard
6 Inkling 2 Small 21.7% Nov 2026 Official leaderboard
7 Grok 4.7 18.3% Sep 2026 Official leaderboard
8 Mistral Large 4 Preview 17.5% Oct 2026 Official leaderboard
9 Gemini 4 Carbon 17.5% Oct 2026 Official leaderboard
10 Mistral Large 4 15.8% Oct 2026 Official leaderboard
11 Qwen3.8 Max (0902) 15.4% Sep 2026 Official leaderboard
12 GLM-5.3 14.6% Aug 2026 Official leaderboard
13 DeepSeek V4.1 Flash 1017 14.2% Oct 2026 Official leaderboard
14 MiMo-V2.6 Pro 13.8% Sep 2026 Official leaderboard
15 MiMo-V2.6 Flash 13.3% Sep 2026 Official leaderboard
16 Kimi K3 12.9% Jul 2026 Official leaderboard
17 Step 5 Preview 12.9% Sep 2026 Official leaderboard
18 Tencent Hy4 12.5% Nov 2026 Official leaderboard
19 Gemini 3.8 Flash 12.1% Sep 2026 Official leaderboard
20 Kimi K3.1 Pro 11.7% Oct 2026 Official leaderboard
21 Gemini 4 Hydrogen 11.3% Nov 2026 Official leaderboard
22 DeepSeek V4 Pro 0813 10.8% Aug 2026 Official leaderboard
23 GLM-5.3-Flash 10.4% Aug 2026 Official leaderboard
24 Claude Sonnet 5.5 10.4% Sep 2026 Official leaderboard
25 Claude Opus 5.5 9.17% Sep 2026 Official leaderboard
26 Tencent Hy4 preview 8.75% Aug 2026 Official leaderboard
27 MiniMax-M3.1-Flash 7.50% Oct 2026 Official leaderboard
28 GPT-6 Astra 7.08% Sep 2026 Official leaderboard
29 GPT-6.1 Sol 6.25% Sep 2026 Official leaderboard
30 Mercury 3 5.00% Oct 2026 Official leaderboard
31 Claude Fable 5.2 5.00% Nov 2026 Official leaderboard
32 Beam 1.0 4.58% Oct 2026 Official leaderboard
33 GPT-6 Luna 4.17% Sep 2026 Official leaderboard
34 Claude Haiku 5.5 4.17% Oct 2026 Official leaderboard
35 Inkling 3.75% Jul 2026 Official leaderboard
36 Nemotron 3.5 Super VL 3.75% Oct 2026 Official leaderboard
37 Inkling-Small 2.92% Jul 2026 Official leaderboard
38 GPT-6.1 Terra 1.67% Oct 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 4 Argon 27.1% Sep 2026 Official leaderboard
2 Muse Spark Open 25% Oct 2026 Official leaderboard
3 Muse Ember 25% Nov 2026 Official leaderboard
4 Muse Wonder 23.8% Oct 2026 Official leaderboard
5 Muse Wonder Full 23.3% Oct 2026 Official leaderboard
6 Inkling 2 Small 21.7% Nov 2026 Lab model card
7 Grok 4.7 18.3% Sep 2026 Official leaderboard
8 Mistral Large 4 Preview 17.5% Oct 2026 Official leaderboard
9 Gemini 4 Carbon 17.5% Oct 2026 Official leaderboard
10 Mistral Large 4 15.8% Oct 2026 Official leaderboard
11 Qwen3.8 Max (0902) 15.4% Sep 2026 Official leaderboard
12 GLM-5.3 14.6% Aug 2026 Official leaderboard
13 DeepSeek V4.1 Flash 1017 14.2% Oct 2026 Official leaderboard
14 MiMo-V2.6 Pro 13.8% Sep 2026 Official leaderboard
15 MiMo-V2.6 Flash 13.3% Sep 2026 Official leaderboard
16 Kimi K3 12.9% Jul 2026 Official leaderboard
17 Step 5 Preview 12.9% Sep 2026 Official leaderboard
18 Tencent Hy4 12.5% Nov 2026 Official leaderboard
19 Gemini 3.8 Flash 12.1% Sep 2026 Official leaderboard
20 Kimi K3.1 Pro 11.7% Oct 2026 Official leaderboard
21 Gemini 4 Hydrogen 11.3% Nov 2026 Official leaderboard
22 DeepSeek V4 Pro 0813 10.8% Aug 2026 Official leaderboard
23 GLM-5.3-Flash 10.4% Aug 2026 Official leaderboard
24 Claude Sonnet 5.5 10.4% Sep 2026 Official leaderboard
25 Claude Opus 5.5 9.17% Sep 2026 Official leaderboard
26 Tencent Hy4 preview 8.75% Aug 2026 Official leaderboard
27 MiniMax-M3.1-Flash 7.50% Oct 2026 Official leaderboard
28 GPT-6 Astra 7.08% Sep 2026 Official leaderboard
29 GPT-6.1 Sol 6.25% Sep 2026 Official leaderboard
30 Mercury 3 5.00% Oct 2026 Official leaderboard
31 Claude Fable 5.2 5.00% Nov 2026 Official leaderboard
32 Beam 1.0 4.58% Oct 2026 Official leaderboard
33 GPT-6 Luna 4.17% Sep 2026 Official leaderboard
34 Claude Haiku 5.5 4.17% Oct 2026 Official leaderboard
35 Inkling 3.75% Jul 2026 Official leaderboard
36 Nemotron 3.5 Super VL 3.75% Oct 2026 Official leaderboard
37 Inkling-Small 2.92% Jul 2026 Official leaderboard
38 GPT-6.1 Terra 1.67% Oct 2026 Official leaderboard

Labs report scores for 1 model in their cards. Switch the scores above to rank them together.

The best score, 27.1%, costs $11 a task

Runs on the cost frontier score higher than any run that's cheaper. Muse Ember comes within 2.1 points of the best for $0.12 a task, 88 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Xiaomi Corp Google DeepMind Meta Cost frontier One model at each effort level
0% 10% 20% 30% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Argon Muse Ember MiMo-V2.6 Flash 0% 10% 20% 30% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Argon Muse Ember

Source: Official Harvey LAB v1.1 (HLAB-Vals-1.1) leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 4 Argon 27.1% $11 153,535
Muse Spark Open 25% $14 115,480
Muse Ember 25% $0.12 95,364
Muse Wonder 23.8% $7.08 100,926
Muse Wonder Full 23.3% $5.41 112,495
Inkling 2 Small 21.7% $1.23 170,530
Grok 4.7 18.3% $11 171,242
Mistral Large 4 Preview 17.5% $2.55 106,618
Gemini 4 Carbon 17.5% $6.52 136,642
Mistral Large 4 15.8% $0.73 63,338
Qwen3.8 Max (0902) 15.4% $3.36 124,007
GLM-5.3 14.6% $3.99 135,019
DeepSeek V4.1 Flash 1017 14.2% $0.61 215,683
MiMo-V2.6 Pro 13.8% $0.22 86,713
MiMo-V2.6 Flash 13.3% $0.09 103,528
Kimi K3 12.9% $4.14 83,561
Step 5 Preview 12.9% $1.13 118,439
Tencent Hy4 12.5% $1.04 100,731
Gemini 3.8 Flash 12.1% $3.66 137,535
Kimi K3.1 Pro 11.7% $9.00 178,849
Gemini 4 Hydrogen 11.3% $3.28 142,480
DeepSeek V4 Pro 0813 10.8% $0.80 78,204
GLM-5.3-Flash 10.4% $0.32 82,011
Claude Sonnet 5.5 10.4% $18 595,872
Claude Opus 5.5 9.17% $22 374,040
Tencent Hy4 preview 8.75% $0.98 104,034
MiniMax-M3.1-Flash 7.50% $0.47 167,645
GPT-6 Astra 7.08% $25 86,677
GPT-6.1 Sol 6.25% $3.54 92,990
Mercury 3 5.00% $0.19 50,774
Claude Fable 5.2 5.00% $19 197,098
Beam 1.0 4.58% 36,042
GPT-6 Luna 4.17% $0.30 103,034
Claude Haiku 5.5 4.17% $4.26 576,846
Inkling 3.75% $1.25 65,114
Nemotron 3.5 Super VL 3.75% $0.37 54,265
Inkling-Small 2.92% $0.18 48,354
GPT-6.1 Terra 1.67% $1.48 61,210

Number one over time Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind
Own leaderboard 1 leader
Gemini 4 Argon 27.1%

Source: Official Harvey LAB v1.1 (HLAB-Vals-1.1) leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon Grok 4.7 GLM-5.3 Inkling 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon Grok 4.7

Sources: Official Harvey LAB v1.1 (HLAB-Vals-1.1) leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Thinking Machines
DateLaunch postScoreAs the post writes it
Nov 5, 2026 Inkling 2 Small 21.7 Harvey LAB v1.1, 21.7%

One lab claim next to the leaderboard

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 1 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Inkling 2 Small 21.7% Harvey LAB v1.1, 21.7% 21.7% Leaderboard Same

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

Anthropic xAI Meta Google DeepMind
Harvey LAB v1.1 (HLAB-Vals-1.1) 7 leaders
Gemini 4 Argon 27.1%
Harvey LAB (Legal Agent Benchmark) 6 leaders
Muse Spark 1.2 25.4%
Harvey LAB v1.1 (HLAB-Vals-1.1) 1 leader
Gemini 4 Argon 27.1%

Source: Official leaderboards

Replaces Harvey LAB (Legal Agent Benchmark)

More in Law, finance and medicine

All 179 in Law, finance and medicine
  • MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy. Qwen 3.8 Max 80.4%
  • HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
  • HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy. Ling-3.1-flash 81.7%
  • KBL In use Models answer Korean legal knowledge and reasoning tasks and Korean bar exam questions. It has 7 knowledge tasks, 4 reasoning tasks and 2,510 bar exam items. Solar Pro 4 77.5%
  • KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. Solar Pro 4 93.2%