MentalBench (Psychiatric Diagnosis)
Models select psychiatric diagnoses for 24,750 synthetic cases. Exact matching requires both diagnoses when the case lacks the evidence needed to distinguish them.
The authors publish accuracy for 22 models across four case types. The dataset has 40 Hugging Face likes.
- Made by
- KAIST and colleagues
- Released
- February 2026
- Kind
- Medicine
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Subject
- Clinical medicine
- Language
- English
- Interaction
- Single turn
- Items from
- Generated
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Biology and medicine
- Strongest model tested
- GPT-5.1 #66 on the capability index
- Last model added
- November 2025
- Scores refreshed
- November 20th, 2026
Claude Sonnet 4.5 leads with 62.7% among models released by November 2025 Claude Sonnet 4.5 leads with 62.7% among models released by November 2025
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. 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 November 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 62.7 Claude Sonnet 4.5
- 62.2 GPT-5.1
- 60.9 Claude Haiku 4.5
- 60.6 GPT-5 mini
- 59.6
Gemini 2.5 Pro
- 57.9
Gemini 2.5 Flash
- 52.4 GPT-4o
- 47.1
Llama 3.1-70B
- 47.1
Qwen2.5-32B
- 47.1 Qwen 3 235B
- 46.5
Qwen2.5-14B
- 44.9
Qwen3-32B
- 41.7
Qwen2.5-72B
- 38.5
Gemma 3 12B
- 38.1
Qwen3-14B
- 36.9
Gemma 3 27B
- 36.1
Qwen2.5-7B
- 33.6
Llama 3.1-8B
- 33.3
Gemma 3 4B
- 33.1
Qwen3-8B
- 23.4 MentaLLaMA 13B
- 19.6 MentaLLaMA 7B
- 1 Claude Sonnet 4.5 62.7%
- 2 GPT-5.1 62.2%
- 3 Claude Haiku 4.5 60.9%
- 4 GPT-5 mini 60.6%
- 5
Gemini 2.5 Pro Jun 2025 59.6%
- 6
Gemini 2.5 Flash Sep 2025 57.9%
- 7 GPT-4o Mar 2025 52.4%
- 8
Llama 3.1-70B 47.1%
- 9
Qwen2.5-32B 47.1%
- 10 Qwen 3 235B 47.1%
- 11
Qwen2.5-14B 46.5%
- 12
Qwen3-32B 44.9%
- 13
Qwen2.5-72B 41.7%
- 14
Gemma 3 12B 38.5%
- 15
Qwen3-14B 38.1%
- 16
Gemma 3 27B 36.9%
- 17
Qwen2.5-7B 36.1%
- 18
Llama 3.1-8B 33.6%
- 19
Gemma 3 4B 33.3%
- 20
Qwen3-8B 33.1%
- 21 MentaLLaMA 13B 23.4%
- 22 MentaLLaMA 7B 19.6%
Source: Official MentalBench (Psychiatric Diagnosis) leaderboard
- 62.7 Claude Sonnet 4.5
- 62.2 GPT-5.1
- 60.9 Claude Haiku 4.5
- 60.6 GPT-5 mini
- 59.6
Gemini 2.5 Pro
- 57.9
Gemini 2.5 Flash
- 52.4 GPT-4o
- 47.1
Llama 3.1-70B
- 47.1
Qwen2.5-32B
- 46.5
Qwen2.5-14B
- 44.9
Qwen3-32B
- 41.7
Qwen2.5-72B
- 38.5
Gemma 3 12B
- 38.1
Qwen3-14B
- 36.9
Gemma 3 27B
- 36.1
Qwen2.5-7B
- 33.6
Llama 3.1-8B
- 33.3
Gemma 3 4B
- 33.1
Qwen3-8B
- 1 Claude Sonnet 4.5 62.7%
- 2 GPT-5.1 62.2%
- 3 Claude Haiku 4.5 60.9%
- 4 GPT-5 mini 60.6%
- 5
Gemini 2.5 Pro Jun 2025 59.6%
- 6
Gemini 2.5 Flash Sep 2025 57.9%
- 7 GPT-4o Mar 2025 52.4%
- 8
Llama 3.1-70B 47.1%
- 9
Qwen2.5-32B 47.1%
- 10
Qwen2.5-14B 46.5%
- 11
Qwen3-32B 44.9%
- 12
Qwen2.5-72B 41.7%
- 13
Gemma 3 12B 38.5%
- 14
Qwen3-14B 38.1%
- 15
Gemma 3 27B 36.9%
- 16
Qwen2.5-7B 36.1%
- 17
Llama 3.1-8B 33.6%
- 18
Gemma 3 4B 33.3%
- 19
Qwen3-8B 33.1%
Source: Official MentalBench (Psychiatric Diagnosis) leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 62.7% | Sep 2025 | Official leaderboard | |
| 2 | GPT-5.1 | 62.2% | Nov 2025 | Official leaderboard | |
| 3 | Claude Haiku 4.5 | 60.9% | Oct 2025 | Official leaderboard | |
| 4 | GPT-5 mini | 60.6% | Aug 2025 | Official leaderboard | |
| 5 | | 59.6% | Jun 2025 | Official leaderboard | |
| 6 | | 57.9% | Sep 2025 | Official leaderboard | |
| 7 | GPT-4o Mar 2025 | 52.4% | Mar 2025 | Official leaderboard | |
| 8 | | 47.1% | Jul 2024 | Official leaderboard | |
| 9 | | 47.1% | Sep 2024 | Official leaderboard | |
| 10 | Qwen 3 235B | 47.1% | Official leaderboard | ||
| 11 | | 46.5% | Sep 2024 | Official leaderboard | |
| 12 | | 44.9% | Apr 2025 | Official leaderboard | |
| 13 | | 41.7% | Sep 2024 | Official leaderboard | |
| 14 | | 38.5% | Mar 2025 | Official leaderboard | |
| 15 | | 38.1% | Apr 2025 | Official leaderboard | |
| 16 | | 36.9% | Mar 2025 | Official leaderboard | |
| 17 | | 36.1% | Sep 2024 | Official leaderboard | |
| 18 | | 33.6% | Jul 2024 | Official leaderboard | |
| 19 | | 33.3% | Mar 2025 | Official leaderboard | |
| 20 | | 33.1% | Apr 2025 | Official leaderboard | |
| 21 | MentaLLaMA 13B | 23.4% | Official leaderboard | ||
| 22 | MentaLLaMA 7B | 19.6% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 62.7% | Sep 2025 | Official leaderboard | |
| 2 | GPT-5.1 | 62.2% | Nov 2025 | Official leaderboard | |
| 3 | Claude Haiku 4.5 | 60.9% | Oct 2025 | Official leaderboard | |
| 4 | GPT-5 mini | 60.6% | Aug 2025 | Official leaderboard | |
| 5 | | 59.6% | Jun 2025 | Official leaderboard | |
| 6 | | 57.9% | Sep 2025 | Official leaderboard | |
| 7 | GPT-4o Mar 2025 | 52.4% | Mar 2025 | Official leaderboard | |
| 8 | | 47.1% | Jul 2024 | Official leaderboard | |
| 9 | | 47.1% | Sep 2024 | Official leaderboard | |
| 10 | | 46.5% | Sep 2024 | Official leaderboard | |
| 11 | | 44.9% | Apr 2025 | Official leaderboard | |
| 12 | | 41.7% | Sep 2024 | Official leaderboard | |
| 13 | | 38.5% | Mar 2025 | Official leaderboard | |
| 14 | | 38.1% | Apr 2025 | Official leaderboard | |
| 15 | | 36.9% | Mar 2025 | Official leaderboard | |
| 16 | | 36.1% | Sep 2024 | Official leaderboard | |
| 17 | | 33.6% | Jul 2024 | Official leaderboard | |
| 18 | | 33.3% | Mar 2025 | Official leaderboard | |
| 19 | | 33.1% | Apr 2025 | Official leaderboard |
Rankings by part
We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Type 1, Type 2 and Type 3. GPT-5.1 leads 1 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Type 1
19 models we track have a score here.
- 1 GPT-5.1 ↑1 94.4%
- 2 Claude Sonnet 4.5 ↓1 94.3%
- 3 Claude Haiku 4.5 94.2%
- 4
Qwen2.5-14B ↑6 94.2%
- 5
Gemini 2.5 Pro Jun 2025 93.9%
- 6
Qwen2.5-72B ↑6 93.8%
- 7
Gemini 2.5 Flash Sep 2025 ↓1 93.4%
- 8
Qwen2.5-32B ↑1 92.9%
- 9
Gemma 3 27B ↑6 92.9%
- 10
Gemma 3 12B ↑3 92.8%
Type 2
19 models we track have a score here.
- 1 Claude Sonnet 4.5 91.2%
- 2 GPT-5.1 90.4%
- 3 GPT-5 mini ↑1 88.5%
- 4
Gemini 2.5 Pro Jun 2025 ↑1 88.2%
- 5 Claude Haiku 4.5 ↓2 87.4%
- 6
Gemini 2.5 Flash Sep 2025 85.3%
- 7
Gemma 3 27B ↑8 84.1%
- 8
Qwen3-32B ↑3 83.1%
- 9
Qwen2.5-72B ↑3 83.1%
- 10
Qwen2.5-32B ↓1 83%
Type 3
19 models we track have a score here.
- 1
Qwen3-8B ↑18 55.4%
- 2
Qwen2.5-72B ↑10 52.8%
- 3
Gemma 3 27B ↑12 52.5%
- 4
Qwen3-14B ↑10 52.3%
- 5
Qwen2.5-7B ↑11 50.7%
- 6
Qwen2.5-32B ↑3 49.2%
- 7
Gemma 3 12B ↑6 48.3%
- 8
Qwen3-32B ↑3 48%
- 9
Qwen2.5-14B ↑1 44.8%
- 10 GPT-4o Mar 2025 ↓3 40.3%
Type 4
19 models we track have a score here.
- 1 GPT-5 mini ↑3 77.7%
- 2 Claude Sonnet 4.5 ↓1 74.8%
- 3 GPT-5.1 ↓1 70.3%
- 4 Claude Haiku 4.5 ↓1 65%
- 5
Gemini 2.5 Flash Sep 2025 ↑1 61%
- 6
Gemini 2.5 Pro Jun 2025 ↓1 59.8%
- 7 GPT-4o Mar 2025 48.3%
- 8
Llama 3.1-70B 35%
- 9
Qwen2.5-14B ↑1 31.8%
- 10
Qwen2.5-32B ↓1 30.9%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official MentalBench (Psychiatric Diagnosis) leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official MentalBench (Psychiatric Diagnosis) 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
- 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
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 12 months ago, in November 2025, for GPT-5.1. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (22 entries).
- Pass Models scored 19 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 12 months ago, in November 2025, for GPT-5.1.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.1 (#66).
- Pass Headroom The best is 62.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on kaist-mlplab.github.io: 19 rows for models we track and 3 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.