Benchmark Almanac

TRLawBench

Gemini 3.1 Pro 9.85 leads since February 2026
Not in model cards

Models answer 297 questions on Turkish law in three stages: 97 questions from official ÖSYM legal exams and 200 harder questions across 14 branches of law. Answers are checked against the answer key. A judge model also grades the reasoning from 0 to 10.

The authors say the Stage 2 and 3 questions come from print sources that were never online. Gemini 3.1 Pro grades the reasoning score, so only the correct count is free of judge bias. The board also scores models with YargıPro, which the page does not define. About 43 GitHub stars.

Released
April 2025
Kind
Law
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Strongest model tested
Claude Opus 5 #10 on the capability index
Last model added
August 2026
Scores refreshed
November 12th, 2026
34
Models scored on its own leaderboard

Gemini 3.1 Pro leads with 9.85 among models released by August 2026

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

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

34 models. Scroll sideways for the rest.

  1. 9.85 Gemini 3.1 Pro
  2. 9.65 GPT-5.6 Sol
  3. 9.64 Claude Fable 5
  4. 9.61 Gemini 3.7 Flash
  5. 9.59 Claude Opus 5
  6. 9.58 Gemini 3.5 Flash
  7. 9.50 GPT-5.6 Terra
  8. 9.46 Gemini 3 Flash
  9. 9.44 Qwen 3.8 Max
  10. 9.38 GPT-5.5
  11. 9.38 Grok 4.6
  12. 9.35 Grok 4.5
  13. 9.25 GPT-5.6 Luna
  14. 9.23 Claude Opus 4.8
  15. 9.23 DeepSeek V4 Pro 0813
  16. 9.10 Kimi K3
  17. 9.08 GPT-5.4
  18. 9.03 Claude Opus 4.6
  19. 9.02 Claude Opus 4.7
  20. 8.97 DeepSeek-V4-Pro
  21. 8.93 DeepSeek V4 Flash 0731
  22. 8.92 Claude Sonnet 5
  23. 8.48 Inkling
  24. 8.46 Qwen 3.8 27B
  25. 8.38 Claude Sonnet 4.6
  26. 8.38 Gemma 4 31B IT
  27. 8.11 Gemma 4 26B A4B
  28. 8.02 GLM-5.2
  29. 7.82 Gemma 4 12B
  30. 7.76 GPT-5.4 Mini
  31. 7.50 Gemini 3.1 Flash-Lite
  32. 7.38 Grok 4.20
  33. 6.97 Qwen 3.6 Plus
  34. 4.78 gpt-oss-120b
  1. 1 Gemini 3.1 Pro 9.85
  2. 2 GPT-5.6 Sol 9.65
  3. 3 Claude Fable 5 9.64
  4. 4 Gemini 3.7 Flash 9.61
  5. 5 Claude Opus 5 9.59
  6. 6 Gemini 3.5 Flash 9.58
  7. 7 GPT-5.6 Terra 9.50
  8. 8 Gemini 3 Flash 9.46
  9. 9 Qwen 3.8 Max 9.44
  10. 10 GPT-5.5 9.38
  11. 11 Grok 4.6 9.38
  12. 12 Grok 4.5 9.35
  13. 13 GPT-5.6 Luna 9.25
  14. 14 Claude Opus 4.8 9.23
  15. 15 DeepSeek V4 Pro 0813 9.23
  16. 16 Kimi K3 9.10
  17. 17 GPT-5.4 9.08
  18. 18 Claude Opus 4.6 9.03
  19. 19 Claude Opus 4.7 9.02
  20. 20 DeepSeek-V4-Pro 8.97
  21. 21 DeepSeek V4 Flash 0731 8.93
  22. 22 Claude Sonnet 5 8.92
  23. 23 Inkling 8.48
  24. 24 Qwen 3.8 27B 8.46
  25. 25 Claude Sonnet 4.6 8.38
  26. 26 Gemma 4 31B IT 8.38
  27. 27 Gemma 4 26B A4B 8.11
  28. 28 GLM-5.2 8.02
  29. 29 Gemma 4 12B 7.82
  30. 30 GPT-5.4 Mini 7.76
  31. 31 Gemini 3.1 Flash-Lite 7.50
  32. 32 Grok 4.20 7.38
  33. 33 Qwen 3.6 Plus 6.97
  34. 34 gpt-oss-120b 4.78

Source: Official TRLawBench leaderboard

#ModelScoreSettingDateSource
1 Gemini 3.1 Pro 9.85 YargıPro Feb 2026 Official leaderboard
2 GPT-5.6 Sol 9.65 Jul 2026 Official leaderboard
3 Claude Fable 5 9.64 YargıPro Jun 2026 Official leaderboard
4 Gemini 3.7 Flash 9.61 YargıPro Aug 2026 Official leaderboard
5 Claude Opus 5 9.59 YargıPro Jul 2026 Official leaderboard
6 Gemini 3.5 Flash 9.58 YargıPro May 2026 Official leaderboard
7 GPT-5.6 Terra 9.50 YargıPro Jul 2026 Official leaderboard
8 Gemini 3 Flash 9.46 Dec 2025 Official leaderboard
9 Qwen 3.8 Max 9.44 YargıPro Aug 2026 Official leaderboard
10 GPT-5.5 9.38 YargıPro Apr 2026 Official leaderboard
11 Grok 4.6 9.38 YargıPro Aug 2026 Official leaderboard
12 Grok 4.5 9.35 YargıPro Jul 2026 Official leaderboard
13 GPT-5.6 Luna 9.25 YargıPro Jul 2026 Official leaderboard
14 Claude Opus 4.8 9.23 YargıPro May 2026 Official leaderboard
15 DeepSeek V4 Pro 0813 9.23 YargıPro Aug 2026 Official leaderboard
16 Kimi K3 9.10 YargıPro Jul 2026 Official leaderboard
17 GPT-5.4 9.08 Mar 2026 Official leaderboard
18 Claude Opus 4.6 9.03 Feb 2026 Official leaderboard
19 Claude Opus 4.7 9.02 Apr 2026 Official leaderboard
20 DeepSeek-V4-Pro 8.97 YargıPro Apr 2026 Official leaderboard
21 DeepSeek V4 Flash 0731 8.93 YargıPro Jul 2026 Official leaderboard
22 Claude Sonnet 5 8.92 YargıPro Jun 2026 Official leaderboard
23 Inkling 8.48 YargıPro Jul 2026 Official leaderboard
24 Qwen 3.8 27B 8.46 YargıPro Aug 2026 Official leaderboard
25 Claude Sonnet 4.6 8.38 Feb 2026 Official leaderboard
26 Gemma 4 31B IT 8.38 MCP tools Apr 2026 Official leaderboard
27 Gemma 4 26B A4B 8.11 MCP tools Apr 2026 Official leaderboard
28 GLM-5.2 8.02 YargıPro Jun 2026 Official leaderboard
29 Gemma 4 12B 7.82 YargıPro Jun 2026 Official leaderboard
30 GPT-5.4 Mini 7.76 Mar 2026 Official leaderboard
31 Gemini 3.1 Flash-Lite 7.50 Mar 2026 Official leaderboard
32 Grok 4.20 7.38 Feb 2026 Official leaderboard
33 Qwen 3.6 Plus 6.97 Mar 2026 Official leaderboard
34 gpt-oss-120b 4.78 Aug 2025 Official leaderboard

3 models have held number one on the official leaderboard since its release in April 2025

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

OpenAI Google DeepMind
Own leaderboard Not tested since Aug 2026
Gemini 3.1 Pro 9.85

Source: Official TRLawBench 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
4.00 6.00 8.00 10.0 2026 4.00 6.00 8.00 10.0 2026

Source: Official TRLawBench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this 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%