Benchmark Almanac

OAB-Bench

Gemini 3.1 Pro 9.39 leads since February 2026
Not in model cards

Models write answers for the second phase of the Brazilian bar examination. The port averages model judges’ scores against grading guidelines.

In April 2025, Claude 3.5 Sonnet led with 7.93 out of 10 and passed all 21 exams. The hub separately reports GPT-4.1 mini on five examples.

Made by
Maritaca AI
Released
April 2025
Kind
Law
Level
Expert
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Capability
Subject
Law
Language
Portuguese
In and out
Text → Text
Interaction
Single turn
Items from
Exams and contests
Refresh
One fixed set
Run by
Anyone
Occupations
Legal
In views
Law and finance
Strongest model tested
Claude Opus 4.6 #39 on the capability index
Last model added
March 2026
Scores refreshed
November 20th, 2026
12
Entries on its own leaderboard, 8 models we track

Gemini 3.1 Pro leads with 9.39 among models released by March 2026 Gemini 3.1 Pro leads with 9.39 among models released by March 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

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

  1. 1 Gemini 3.1 Pro GPT-5.2 judge, structured 9.39
  2. 2 Claude Opus 4.6 GPT-5.2 judge, structured 8.82
  3. 3 GPT-5.2 GPT-5.2 judge, structured 8.66
  4. 4 Claude Sonnet 4.6 GPT-5.2 judge, structured 8.27
  5. 5 Sabiá-4 GPT-5.2 judge, structured 7.96
  6. 6 Gemini 3.1 Flash-Lite GPT-5.2 judge, structured 7.60
  7. 7 Sabiazinho-4 GPT-5.2 judge, non-structured 6.92
  8. 8 Sabiá-3.1 GPT-5.2 judge, structured 6.91
  9. 9 Qwen3.5 397B-A17B GPT-5.2 judge, structured 6.69
  10. 10 GPT-5 mini GPT-5.2 judge, structured 6.52
  11. 11 Qwen3.5-35B-A3B GPT-5.2 judge, non-structured 6.18
  12. 12 Sabiazinho-3 GPT-5.2 judge, structured 6.12

Source: Official OAB-Bench leaderboard

#EntryScoreSettingDateSource
1 Gemini 3.1 Pro 9.39 GPT-5.2 judge, structured Feb 2026 Official leaderboard
2 Claude Opus 4.6 8.82 GPT-5.2 judge, structured Feb 2026 Official leaderboard
3 GPT-5.2 8.66 GPT-5.2 judge, structured Dec 2025 Official leaderboard
4 Claude Sonnet 4.6 8.27 GPT-5.2 judge, structured Feb 2026 Official leaderboard
5 Sabiá-4 7.96 GPT-5.2 judge, structured Official leaderboard
6 Gemini 3.1 Flash-Lite 7.60 GPT-5.2 judge, structured Mar 2026 Official leaderboard
7 Sabiazinho-4 6.92 GPT-5.2 judge, non-structured Official leaderboard
8 Sabiá-3.1 6.91 GPT-5.2 judge, structured Official leaderboard
9 Qwen3.5 397B-A17B 6.69 GPT-5.2 judge, structured Feb 2026 Official leaderboard
10 GPT-5 mini 6.52 GPT-5.2 judge, structured Aug 2025 Official leaderboard
11 Qwen3.5-35B-A3B 6.18 GPT-5.2 judge, non-structured Feb 2026 Official leaderboard
12 Sabiazinho-3 6.12 GPT-5.2 judge, structured Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 3.1 Pro 9.39 GPT-5.2 judge, structured Feb 2026 Official leaderboard
2 Claude Opus 4.6 8.82 GPT-5.2 judge, structured Feb 2026 Official leaderboard
3 GPT-5.2 8.66 GPT-5.2 judge, structured Dec 2025 Official leaderboard
4 Claude Sonnet 4.6 8.27 GPT-5.2 judge, structured Feb 2026 Official leaderboard
5 Gemini 3.1 Flash-Lite 7.60 GPT-5.2 judge, structured Mar 2026 Official leaderboard
6 Qwen3.5 397B-A17B 6.69 GPT-5.2 judge, structured Feb 2026 Official leaderboard
7 GPT-5 mini 6.52 GPT-5.2 judge, structured Aug 2025 Official leaderboard
8 Qwen3.5-35B-A3B 6.18 GPT-5.2 judge, non-structured Feb 2026 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemini 3.1 Pro
GPT-5.2 jud…
9.39
Claude Opus 4.6
GPT-5.2 jud…
8.82
GPT-5.2
GPT-5.2 jud…
8.66
Claude Sonnet 4.6
GPT-5.2 jud…
8.27
Gemini 3.1 Flash-Lite
GPT-5.2 jud…
7.60
Qwen3.5 397B-A17B
GPT-5.2 jud…
6.69
GPT-5 mini
GPT-5.2 jud…
6.52
Qwen3.5-35B-A3B
GPT-5.2 jud…
6.18

Source: Official OAB-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.1 Pro GPT-5.2 judge, structured 9.39
Gemini 3.1 Pro GPT-5.2 judge, non-structured 9.34
Claude Opus 4.6 GPT-5.2 judge, structured 8.82
Claude Opus 4.6 GPT-5.2 judge, non-structured 8.80
GPT-5.2 GPT-5.2 judge, structured 8.66
GPT-5.2 GPT-5.2 judge, non-structured 8.61
Claude Sonnet 4.6 GPT-5.2 judge, structured 8.27
Claude Sonnet 4.6 GPT-5.2 judge, non-structured 8.23
Gemini 3.1 Flash-Lite GPT-5.2 judge, structured 7.60
Gemini 3.1 Flash-Lite GPT-5.2 judge, non-structured 7.57
Qwen3.5 397B-A17B GPT-5.2 judge, structured 6.69
Qwen3.5 397B-A17B GPT-5.2 judge, non-structured 6.67
GPT-5 mini GPT-5.2 judge, structured 6.52
GPT-5 mini GPT-5.2 judge, non-structured 6.50
Qwen3.5-35B-A3B GPT-5.2 judge, non-structured 6.18
Qwen3.5-35B-A3B GPT-5.2 judge, structured 6.17

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemini 3.1 Pro
GPT-5.2 jud…
9.39
Claude Opus 4.6
GPT-5.2 jud…
8.82
GPT-5.2
GPT-5.2 jud…
8.66
Claude Sonnet 4.6
GPT-5.2 jud…
8.27
Gemini 3.1 Flash-Lite
GPT-5.2 jud…
7.60
Qwen3.5 397B-A17B
GPT-5.2 jud…
6.69
GPT-5 mini
GPT-5.2 jud…
6.52
Qwen3.5-35B-A3B
GPT-5.2 jud…
6.18

Source: Official OAB-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.1 Pro GPT-5.2 judge, structured 9.39
Gemini 3.1 Pro GPT-5.2 judge, non-structured 9.34
Claude Opus 4.6 GPT-5.2 judge, structured 8.82
Claude Opus 4.6 GPT-5.2 judge, non-structured 8.80
GPT-5.2 GPT-5.2 judge, structured 8.66
GPT-5.2 GPT-5.2 judge, non-structured 8.61
Claude Sonnet 4.6 GPT-5.2 judge, structured 8.27
Claude Sonnet 4.6 GPT-5.2 judge, non-structured 8.23
Gemini 3.1 Flash-Lite GPT-5.2 judge, structured 7.60
Gemini 3.1 Flash-Lite GPT-5.2 judge, non-structured 7.57
Qwen3.5 397B-A17B GPT-5.2 judge, structured 6.69
Qwen3.5 397B-A17B GPT-5.2 judge, non-structured 6.67
GPT-5 mini GPT-5.2 judge, structured 6.52
GPT-5 mini GPT-5.2 judge, non-structured 6.50
Qwen3.5-35B-A3B GPT-5.2 judge, non-structured 6.18
Qwen3.5-35B-A3B GPT-5.2 judge, structured 6.17

4 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 Anthropic Google DeepMind
Own leaderboard Not tested since Mar 2026

Source: Official OAB-Bench leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard
6.00 7.00 8.00 9.00 10.0 2026 6.00 7.00 8.00 9.00 10.0 2026

Source: Official OAB-Bench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Law, finance and medicine

All 637 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%
  • 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. o1-preview 92.7%
  • Harvey LAB-AA v1.1 In use Artificial Analysis's run of Harvey's Legal Agent Benchmark on Harvey's newer private set of 120 tasks across 24 practice areas. A deliverable with a material hallucination scores zero, and the headline is the share of judges who find every rubric criterion passed. GPT-6.2 Astra 15%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 8 months ago, in March 2026, for Gemini 3.1 Flash-Lite. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (12 entries).
  • Pass Models scored 8 models we track, plus 4 other entries on its leaderboard.
  • Warning Newest score The newest came 8 months ago, in March 2026, for Gemini 3.1 Flash-Lite.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 4.6 (#39).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 16 rows for models we track and 4 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check