Benchmark Almanac

CaseLaw v2

Grok 4.3 79.3% leads since Apr 2026 Not in model cards

Models answer private questions over recent Canadian court decisions, such as finding key cases and answering across several documents.

Replaced the original CaseLaw benchmark after models neared 90% on it. Updated 2026-03-02 and archived as saturated, last run 2026-05-04 with 54 models.

Made by Vals AI (with Jurisage, Together AI and Stanford CodeX) Released Aug 2025 Law Tasks kept private Last model added Apr 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
Open-ended Free-form text, documents, designs or media judged as a whole.
Leaderboard
53
Models scored on its own leaderboard

Grok 4.3 leads with 79.3% among models released by Apr 2026

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.

The newest model with a score here came out in Apr 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

53 models. Scroll sideways for the rest.

  1. 79.3 Grok 4.3
  2. 73.4 GPT-5.1
  3. 69.9 GPT-4.1
  4. 68.5 GPT-5 mini
  5. 68.4 Claude Opus 4.7
  6. 66.5 GPT-5
  7. 66.2 GPT-5.5
  8. 66.0 GPT-5.2
  9. 65.8 Grok 4
  10. 65.7 Grok 4 Fast
  11. 65.7 Kimi K2 Thinking
  12. 64.8 Gemini 3.1 Pro
  13. 64.5 Cohere Command A
  14. 64.0 Claude Sonnet 4.6
  15. 63.9 Gemini 2.5 Pro
  16. 63.8 GPT-5.4
  17. 63.1 Muse Spark
  18. 62.6 Claude Opus 4.5
  19. 62.2 Claude Sonnet 4.5
  20. 62.1 Claude Opus 4.6
  21. 61.4 Mistral Large 3
  22. 61.2 Kimi K2.6
  23. 60.9 MiniMax-M2.7
  24. 60.5 Grok 4.1 Fast
  25. 59.7 GPT-4o
  26. 59.7 Qwen 3.5 Plus (hosted 397B-A17B)
  27. 59.4 DeepSeek-V4-Pro
  28. 58.7 Kimi K2.5
  29. 57.9 Trinity Large Thinking
  30. 56.5 Claude Haiku 4.5
  31. 56.0 Qwen 3.5 Flash (hosted 35B-A3B)
  32. 55.8 Gemini 3 Flash
  33. 55.8 MiniMax-M2.1
  34. 55.4 DeepSeek-V3.2
  35. 55.0 Qwen3-Max
  36. 55.0 Gemini 3.1 Flash-Lite
  37. 54.9 GLM-4.7
  38. 54.5 Grok 4.20
  39. 53.9 DeepSeek-V3.1
  40. 53.5 MiniMax-M2.5
  41. 53.2 Qwen3.6 27B
  42. 53.0 Gemini 3 Pro
  43. 52.6 GPT-5 nano
  44. 52.6 Gemma 4 31B IT
  45. 52.5 GLM-5
  46. 51.9 GPT-5.4 Nano
  47. 51.7 GPT-5.4 Mini
  48. 51.5 GLM-5.1
  49. 51.5 Qwen 3.6 Plus
  50. 48.8 gpt-oss-120b
  51. 47.9 Qwen 3.6 Max (Preview)
  52. 44.2 Mistral Medium 3.5
  53. 43.8 gpt-oss-20b
  1. 1 Grok 4.3 79.3%
  2. 2 GPT-5.1 73.4%
  3. 3 GPT-4.1 69.9%
  4. 4 GPT-5 mini 68.5%
  5. 5 Claude Opus 4.7 68.4%
  6. 6 GPT-5 66.5%
  7. 7 GPT-5.5 66.2%
  8. 8 GPT-5.2 66%
  9. 9 Grok 4 65.8%
  10. 10 Grok 4 Fast 65.7%
  11. 11 Kimi K2 Thinking 65.7%
  12. 12 Gemini 3.1 Pro 64.8%
  13. 13 Cohere Command A 64.5%
  14. 14 Claude Sonnet 4.6 64%
  15. 15 Gemini 2.5 Pro Jun 2025 63.9%
  16. 16 GPT-5.4 xhigh 63.8%
  17. 17 Muse Spark 63.1%
  18. 18 Claude Opus 4.5 thinking 62.6%
  19. 19 Claude Sonnet 4.5 thinking 62.2%
  20. 20 Claude Opus 4.6 thinking 62.1%
  21. 21 Mistral Large 3 61.4%
  22. 22 Kimi K2.6 61.2%
  23. 23 MiniMax-M2.7 60.9%
  24. 24 Grok 4.1 Fast 60.5%
  25. 25 GPT-4o Aug 2024 59.7%
  26. 26 Qwen 3.5 Plus (hosted 397B-A17B) thinking 59.7%
  27. 27 DeepSeek-V4-Pro 59.4%
  28. 28 Kimi K2.5 thinking 58.7%
  29. 29 Trinity Large Thinking 57.9%
  30. 30 Claude Haiku 4.5 thinking 56.5%
  31. 31 Qwen 3.5 Flash (hosted 35B-A3B) 56%
  32. 32 Gemini 3 Flash 55.8%
  33. 33 MiniMax-M2.1 55.8%
  34. 34 DeepSeek-V3.2 thinking 55.4%
  35. 35 Qwen3-Max 55%
  36. 36 Gemini 3.1 Flash-Lite 55%
  37. 37 GLM-4.7 54.9%
  38. 38 Grok 4.20 54.5%
  39. 39 DeepSeek-V3.1 53.9%
  40. 40 MiniMax-M2.5 53.5%
  41. 41 Qwen3.6 27B 53.2%
  42. 42 Gemini 3 Pro 53%
  43. 43 GPT-5 nano 52.6%
  44. 44 Gemma 4 31B IT 52.6%
  45. 45 GLM-5 thinking 52.5%
  46. 46 GPT-5.4 Nano 51.9%
  47. 47 GPT-5.4 Mini 51.7%
  48. 48 GLM-5.1 51.5%
  49. 49 Qwen 3.6 Plus 51.5%
  50. 50 gpt-oss-120b 48.8%
  51. 51 Qwen 3.6 Max (Preview) 47.9%
  52. 52 Mistral Medium 3.5 44.2%
  53. 53 gpt-oss-20b 43.8%
#ModelScoreSettingDateSource
1 Grok 4.3 79.3% Apr 2026 Official leaderboard
2 GPT-5.1 73.4% Nov 2025 Official leaderboard
3 GPT-4.1 69.9% Apr 2025 Official leaderboard
4 GPT-5 mini 68.5% Aug 2025 Official leaderboard
5 Claude Opus 4.7 68.4% Apr 2026 Official leaderboard
6 GPT-5 66.5% Aug 2025 Official leaderboard
7 GPT-5.5 66.2% Apr 2026 Official leaderboard
8 GPT-5.2 66% Dec 2025 Official leaderboard
9 Grok 4 65.8% Jul 2025 Official leaderboard
10 Grok 4 Fast 65.7% Sep 2025 Official leaderboard
11 Kimi K2 Thinking 65.7% Nov 2025 Official leaderboard
12 Gemini 3.1 Pro 64.8% Feb 2026 Official leaderboard
13 Cohere Command A 64.5% Mar 2025 Official leaderboard
14 Claude Sonnet 4.6 64% Feb 2026 Official leaderboard
15 Gemini 2.5 Pro Jun 2025 63.9% Jun 2025 Official leaderboard
16 GPT-5.4 63.8% xhigh Mar 2026 Official leaderboard
17 Muse Spark 63.1% Apr 2026 Official leaderboard
18 Claude Opus 4.5 62.6% thinking Nov 2025 Official leaderboard
19 Claude Sonnet 4.5 62.2% thinking Sep 2025 Official leaderboard
20 Claude Opus 4.6 62.1% thinking Feb 2026 Official leaderboard
21 Mistral Large 3 61.4% Dec 2025 Official leaderboard
22 Kimi K2.6 61.2% Apr 2026 Official leaderboard
23 MiniMax-M2.7 60.9% Mar 2026 Official leaderboard
24 Grok 4.1 Fast 60.5% Nov 2025 Official leaderboard
25 GPT-4o Aug 2024 59.7% Aug 2024 Official leaderboard
26 Qwen 3.5 Plus (hosted 397B-A17B) 59.7% thinking Feb 2026 Official leaderboard
27 DeepSeek-V4-Pro 59.4% Apr 2026 Official leaderboard
28 Kimi K2.5 58.7% thinking Jan 2026 Official leaderboard
29 Trinity Large Thinking 57.9% Apr 2026 Official leaderboard
30 Claude Haiku 4.5 56.5% thinking Oct 2025 Official leaderboard
31 Qwen 3.5 Flash (hosted 35B-A3B) 56% Feb 2026 Official leaderboard
32 Gemini 3 Flash 55.8% Dec 2025 Official leaderboard
33 MiniMax-M2.1 55.8% Dec 2025 Official leaderboard
34 DeepSeek-V3.2 55.4% thinking Dec 2025 Official leaderboard
35 Qwen3-Max 55% Sep 2025 Official leaderboard
36 Gemini 3.1 Flash-Lite 55% Mar 2026 Official leaderboard
37 GLM-4.7 54.9% Dec 2025 Official leaderboard
38 Grok 4.20 54.5% Feb 2026 Official leaderboard
39 DeepSeek-V3.1 53.9% Aug 2025 Official leaderboard
40 MiniMax-M2.5 53.5% Feb 2026 Official leaderboard
41 Qwen3.6 27B 53.2% Apr 2026 Official leaderboard
42 Gemini 3 Pro 53% Nov 2025 Official leaderboard
43 GPT-5 nano 52.6% Aug 2025 Official leaderboard
44 Gemma 4 31B IT 52.6% Apr 2026 Official leaderboard
45 GLM-5 52.5% thinking Feb 2026 Official leaderboard
46 GPT-5.4 Nano 51.9% Mar 2026 Official leaderboard
47 GPT-5.4 Mini 51.7% Mar 2026 Official leaderboard
48 GLM-5.1 51.5% Apr 2026 Official leaderboard
49 Qwen 3.6 Plus 51.5% Mar 2026 Official leaderboard
50 gpt-oss-120b 48.8% Aug 2025 Official leaderboard
51 Qwen 3.6 Max (Preview) 47.9% Apr 2026 Official leaderboard
52 Mistral Medium 3.5 44.2% Apr 2026 Official leaderboard
53 gpt-oss-20b 43.8% Aug 2025 Official leaderboard

3 models have held number one since its release in Aug 2025

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

OpenAI xAI
Own leaderboard Not tested since Apr 2026
Grok 4.3 79.3%

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% 2025 2026 Near the ceiling (85%) Grok 4.3 GPT-5.1 GPT-4.1 Cohere Command A GPT-4o 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Grok 4.3 GPT-4.1

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 Law, finance and medicine

All 140 in Law, finance and medicine
  • 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%
  • 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 In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • 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%
  • PRBench In use Model answers 1,100 expert-written law and finance questions. LLM judge grades against weighted expert rubrics of 10-30 criteria each. Muse Spark 1.3 61.6%