Benchmark Almanac

LegalReasonBench

Gemini 4 Argon 50.8% leads since October 2026
Not in model cards

Models work long-horizon legal matters built by practicing lawyers, writing memoranda and court filings across dependent assignments, in closed-record and open-research settings.

Rule and application scores drop 11 to 38 points when models must find the law themselves.

Made by
micro1
Released
October 2026
Kind
Law
Level
Expert
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last model added
November 2026
Scores refreshed
November 12th, 2026
10
Models scored on its own leaderboard

Gemini 4 Argon leads with 50.8%

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. Every run is in the next chart.

  1. 50.8 Gemini 4 Argon
  2. 50.6 Claude Fable 5.1
  3. 49.8 Claude Fable 5.2
  4. 46.5 Claude Opus 5.5
  5. 40.4 Muse Spark 1.3
  6. 40.4 Grok 4.7
  7. 39.6 Gemini 3.8 Flash
  8. 39.5 GPT-6 Astra
  9. 36.4 Kimi K3.1 Pro
  10. 31.9 GPT-6 Sol
  1. 1 Gemini 4 Argon 50.8%
  2. 2 Claude Fable 5.1 50.6%
  3. 3 Claude Fable 5.2 49.8%
  4. 4 Claude Opus 5.5 46.5%
  5. 5 Muse Spark 1.3 40.4%
  6. 6 Grok 4.7 40.4%
  7. 7 Gemini 3.8 Flash 39.6%
  8. 8 GPT-6 Astra 39.5%
  9. 9 Kimi K3.1 Pro 36.4%
  10. 10 GPT-6 Sol 31.9%

Source: Official LegalReasonBench leaderboard

#ModelScoreSettingDateSource
1 Gemini 4 Argon 50.8% Sep 2026 Official leaderboard
2 Claude Fable 5.1 50.6% Sep 2026 Official leaderboard
3 Claude Fable 5.2 49.8% Nov 2026 Official leaderboard
4 Claude Opus 5.5 46.5% Sep 2026 Official leaderboard
5 Muse Spark 1.3 40.4% Sep 2026 Official leaderboard
6 Grok 4.7 40.4% Sep 2026 Official leaderboard
7 Gemini 3.8 Flash 39.6% Sep 2026 Official leaderboard
8 GPT-6 Astra 39.5% Sep 2026 Official leaderboard
9 Kimi K3.1 Pro 36.4% Oct 2026 Official leaderboard
10 GPT-6 Sol 31.9% Sep 2026 Official leaderboard

The best score, 50.8%, costs $6.71 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.

Meta Google DeepMind Cost frontier One model at each effort level
0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Argon Muse Spark 1.3 0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Argon Muse Spark 1.3

Source: Official LegalReasonBench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 4 Argon 50.8% $6.71 169,000
Claude Fable 5.1 50.6% $20 86,000
Claude Fable 5.2 49.8%
Claude Opus 5.5 46.5% $8.08 74,000
Muse Spark 1.3 40.4% $0.87 36,000
Grok 4.7 40.4% $3.42 38,000
Gemini 3.8 Flash 39.6% $3.96 213,000
GPT-6 Astra 39.5% $22 33,000
Kimi K3.1 Pro 36.4%
GPT-6 Sol 31.9% $2.45 23,000

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 50.8%

Source: Official LegalReasonBench 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 Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon

Source: Official LegalReasonBench 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%