Benchmark Almanac

Vals Legal Index

Muse Wonder Full 77.9% leads since Oct 2026 Not in model cards

Agents are scored on Legal Research Bench, HLAB-Vals-1.1 and Tax Agent Bench, mapped to practice areas weighted by share of US law-office revenue.

Practice areas without a covering benchmark (7% of revenue) are not scored. 33 models on 2026-10-29, led by Wonder Full at 77.86%.

Made by Vals AI Released Oct 2026 Composite work indices Tasks kept private Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Leaderboard
33
Models scored on its own leaderboard

Muse Wonder Full leads with 77.9%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

33 models. Scroll sideways for the rest.

  1. 77.9 Muse Wonder Full
  2. 76.7 Gemini 4 Argon
  3. 75.9 Gemini 4 Carbon
  4. 75.3 Muse Wonder
  5. 74.6 GLM-5.3
  6. 73.9 Qwen3.8 Max (0902)
  7. 72.6 Grok 4.7
  8. 72.3 Claude Opus 5.5
  9. 72.1 MiMo-V2.6 Pro
  10. 71.9 DeepSeek V4.1 Flash 1017
  11. 71.7 Kimi K3
  12. 71.7 Tencent Hy4 preview
  13. 71.0 GLM-5.3-Flash
  14. 70.3 Claude Haiku 5.5
  15. 70.2 Step 5 Preview
  16. 69.8 Gemini 3.8 Flash
  17. 69.0 DeepSeek V4 Pro 0813
  18. 69.0 MiMo-V2.6 Flash
  19. 68.7 GPT-6.1 Sol
  20. 68.5 Muse Spark Open
  21. 68.4 GPT-6 Astra
  22. 68.0 Claude Sonnet 5.5
  23. 67.8 Mistral Large 4 (Le Chonk, open weights)
  24. 67.1 MiniMax-M3.1-Flash
  25. 67.0 Kimi K3.1 Pro
  26. 65.6 Mistral Large 4
  27. 63.0 Mercury 3
  28. 62.2 GPT-6 Luna
  29. 61.8 Beam 1.0
  30. 58.8 GPT-6.1 Terra
  31. 57.8 Inkling-Small
  32. 55.4 Nemotron 3.5 Super VL
  33. 54.2 Inkling
  1. 1 Muse Wonder Full 77.9%
  2. 2 Gemini 4 Argon 76.7%
  3. 3 Gemini 4 Carbon 75.9%
  4. 4 Muse Wonder 75.3%
  5. 5 GLM-5.3 74.6%
  6. 6 Qwen3.8 Max (0902) 73.9%
  7. 7 Grok 4.7 72.6%
  8. 8 Claude Opus 5.5 72.3%
  9. 9 MiMo-V2.6 Pro 72.1%
  10. 10 DeepSeek V4.1 Flash 1017 71.9%
  11. 11 Kimi K3 71.7%
  12. 12 Tencent Hy4 preview 71.7%
  13. 13 GLM-5.3-Flash 71%
  14. 14 Claude Haiku 5.5 70.3%
  15. 15 Step 5 Preview 70.2%
  16. 16 Gemini 3.8 Flash 69.8%
  17. 17 DeepSeek V4 Pro 0813 69%
  18. 18 MiMo-V2.6 Flash 69%
  19. 19 GPT-6.1 Sol 68.7%
  20. 20 Muse Spark Open 68.5%
  21. 21 GPT-6 Astra 68.4%
  22. 22 Claude Sonnet 5.5 68%
  23. 23 Mistral Large 4 (Le Chonk, open weights) 67.8%
  24. 24 MiniMax-M3.1-Flash 67.1%
  25. 25 Kimi K3.1 Pro 67%
  26. 26 Mistral Large 4 65.6%
  27. 27 Mercury 3 63%
  28. 28 GPT-6 Luna 62.2%
  29. 29 Beam 1.0 61.8%
  30. 30 GPT-6.1 Terra 58.8%
  31. 31 Inkling-Small 57.8%
  32. 32 Nemotron 3.5 Super VL 55.4%
  33. 33 Inkling 54.2%
#ModelScoreSettingDateSource
1 Muse Wonder Full 77.9% Oct 2026 Official leaderboard
2 Gemini 4 Argon 76.7% Sep 2026 Official leaderboard
3 Gemini 4 Carbon 75.9% Oct 2026 Official leaderboard
4 Muse Wonder 75.3% Oct 2026 Official leaderboard
5 GLM-5.3 74.6% Aug 2026 Official leaderboard
6 Qwen3.8 Max (0902) 73.9% Sep 2026 Official leaderboard
7 Grok 4.7 72.6% Sep 2026 Official leaderboard
8 Claude Opus 5.5 72.3% Sep 2026 Official leaderboard
9 MiMo-V2.6 Pro 72.1% Sep 2026 Official leaderboard
10 DeepSeek V4.1 Flash 1017 71.9% Oct 2026 Official leaderboard
11 Kimi K3 71.7% Jul 2026 Official leaderboard
12 Tencent Hy4 preview 71.7% Aug 2026 Official leaderboard
13 GLM-5.3-Flash 71% Aug 2026 Official leaderboard
14 Claude Haiku 5.5 70.3% Oct 2026 Official leaderboard
15 Step 5 Preview 70.2% Sep 2026 Official leaderboard
16 Gemini 3.8 Flash 69.8% Sep 2026 Official leaderboard
17 DeepSeek V4 Pro 0813 69% Aug 2026 Official leaderboard
18 MiMo-V2.6 Flash 69% Sep 2026 Official leaderboard
19 GPT-6.1 Sol 68.7% Sep 2026 Official leaderboard
20 Muse Spark Open 68.5% Oct 2026 Official leaderboard
21 GPT-6 Astra 68.4% Sep 2026 Official leaderboard
22 Claude Sonnet 5.5 68% Sep 2026 Official leaderboard
23 Mistral Large 4 (Le Chonk, open weights) 67.8% Oct 2026 Official leaderboard
24 MiniMax-M3.1-Flash 67.1% Oct 2026 Official leaderboard
25 Kimi K3.1 Pro 67% Oct 2026 Official leaderboard
26 Mistral Large 4 65.6% Oct 2026 Official leaderboard
27 Mercury 3 63% Oct 2026 Official leaderboard
28 GPT-6 Luna 62.2% Sep 2026 Official leaderboard
29 Beam 1.0 61.8% Oct 2026 Official leaderboard
30 GPT-6.1 Terra 58.8% Oct 2026 Official leaderboard
31 Inkling-Small 57.8% Jul 2026 Official leaderboard
32 Nemotron 3.5 Super VL 55.4% Oct 2026 Official leaderboard
33 Inkling 54.2% Jul 2026 Official leaderboard

The best score, 77.9%, costs $3.72 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Z.ai Xiaomi Corp Meta Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Muse Wonder Full MiMo-V2.6 Pro MiMo-V2.6 Flash 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Muse Wonder Full MiMo-V2.6 Pro
ModelEffortHarnessScoreCost per taskOutput tokens
Muse Wonder Full 77.9% $3.72
Gemini 4 Argon 76.7% $6.82
Gemini 4 Carbon 75.9% $7.20
Muse Wonder 75.3% $4.32
GLM-5.3 74.6% $2.68
Qwen3.8 Max (0902) 73.9% $4.25
Grok 4.7 72.6% $5.82
Claude Opus 5.5 72.3% $21
MiMo-V2.6 Pro 72.1% $0.18
DeepSeek V4.1 Flash 1017 71.9% $0.59
Kimi K3 71.7% $3.47
Tencent Hy4 preview 71.7% $0.81
GLM-5.3-Flash 71% $0.19
Claude Haiku 5.5 70.3% $2.36
Step 5 Preview 70.2% $0.84
Gemini 3.8 Flash 69.8% $2.05
DeepSeek V4 Pro 0813 69% $0.87
MiMo-V2.6 Flash 69% $0.07
GPT-6.1 Sol 68.7% $2.75
Muse Spark Open 68.5% $5.36
GPT-6 Astra 68.4% $14
Claude Sonnet 5.5 68% $14
Mistral Large 4 (Le Chonk, open weights) 67.8% $0.78
MiniMax-M3.1-Flash 67.1% $0.41
Kimi K3.1 Pro 67% $6.63
Mistral Large 4 65.6% $2.22
Mercury 3 63% $0.10
GPT-6 Luna 62.2% $0.31
Beam 1.0 61.8%
GPT-6.1 Terra 58.8% $2.29
Inkling-Small 57.8% $0.14
Nemotron 3.5 Super VL 55.4% $0.16
Inkling 54.2% $0.70

2 models have held number one since its release in Oct 2026

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

Google DeepMind Meta
Own leaderboard 2 leaders
Muse Wonder Full 77.9%

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 Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Muse Wonder Full GLM-5.3 Inkling-Small 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Muse Wonder Full Inkling-Small

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Economic work

All 91 in Economic work
  • JobBench In use 130 tasks from 35 occupations workers most want delegated. Agent works through messy reference files, graded by fact-anchored binary rubrics. GLM-5.3 61.4%
  • GDPval In use Real work tasks from 44 occupations with reference files. Experts blindly compare model deliverables with expert work. Metric is win-or-tie rate. Seed2.1 Pro 87.9%
  • GDPval-AA v2.1 In use Agents with shell and web access complete 220 GDPval gold tasks from 44 occupations. Blind pairwise comparisons give Elo ratings. Claude Opus 5.5 1,846
  • AA-Briefcase v1.1 In use Version 1.1 of Artificial Analysis's multi-hour office work evaluation, with Elo ratings fitted by Crowd-BT. Claude Opus 5.5 1,822
  • CoWorkBench In use Agent completes long-horizon productivity tasks across computer science, finance, law and medicine. Qwen in-house scoring (method unpublished). Qwen3.8-Omni-Flash 75.3%
  • OfficeQA In use Grounded reasoning over a large corpus of U.S. Treasury Bulletin documents: locate tables and compute precise numeric answers. Claude Opus 4.7 86.3%