PSEBench
Models triage patient safety reports under Minnesota rules. They identify reportable events, request missing facts, or abstain when the evidence permits multiple verdicts.
The authors evaluated 15 models on 5,074 cases.
- Made by
- Scale AI and collaborators
- Released
- June 2026
- Kind
- Medicine
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Measures
- Capability
- Subject
- Health sciences, Law
- Language
- English
- Interaction
- Agentic
- Environment
- A simulator
- Items from
- Generated, Real work and use
- Refresh
- One fixed set
- Run by
- A host
- Occupations
- Healthcare practitioners and technical
- In views
- Agents, Robotics, Biology and medicine, Law and finance
- Strongest model tested
- GPT-5.5 #21 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
Gemini 3.1 Pro leads with 95.3% among models released by April 2026 Gemini 3.1 Pro leads with 95.3% among models released by April 2026
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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 April 2026, 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
- 1
Gemini 3.1 Pro 95.3%
- 2 GPT-5.5 94.6%
- 3 Claude Opus 4.7 92.3%
- 4 GPT-5 90.2%
- 5
Gemini 2.5 Flash Sep 2025 85.8%
- 6 Claude Sonnet 4.6 84.5%
- 7
DeepSeek-R1 May 2025 83.7%
- 8 gpt-oss-120b 79%
- 9
Qwen3-235B-A22B 76.7%
- 10 GPT-5 nano 71.9%
- 11 HuatuoGPT-o1 (70B) 61.1%
- 12
Mistral Small 3.2 60.5%
- 13
Llama 3.1-8B 52.4%
- 14 MedGemma (27B) 51.3%
- 15 HuatuoGPT-o1 (8B) 45.5%
Source: The PSEBench paper
- 1
Gemini 3.1 Pro 95.3%
- 2 GPT-5.5 94.6%
- 3 Claude Opus 4.7 92.3%
- 4 GPT-5 90.2%
- 5
Gemini 2.5 Flash Sep 2025 85.8%
- 6 Claude Sonnet 4.6 84.5%
- 7
DeepSeek-R1 May 2025 83.7%
- 8 gpt-oss-120b 79%
- 9
Qwen3-235B-A22B 76.7%
- 10 GPT-5 nano 71.9%
- 11
Mistral Small 3.2 60.5%
- 12
Llama 3.1-8B 52.4%
Source: The PSEBench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 95.3% | Feb 2026 | Official leaderboard | |
| 2 | GPT-5.5 | 94.6% | Apr 2026 | Official leaderboard | |
| 3 | Claude Opus 4.7 | 92.3% | Apr 2026 | Official leaderboard | |
| 4 | GPT-5 | 90.2% | Aug 2025 | Official leaderboard | |
| 5 | | 85.8% | Sep 2025 | Official leaderboard | |
| 6 | Claude Sonnet 4.6 | 84.5% | Feb 2026 | Official leaderboard | |
| 7 | | 83.7% | May 2025 | Official leaderboard | |
| 8 | gpt-oss-120b | 79% | Aug 2025 | Official leaderboard | |
| 9 | | 76.7% | Apr 2025 | Official leaderboard | |
| 10 | GPT-5 nano | 71.9% | Aug 2025 | Official leaderboard | |
| 11 | HuatuoGPT-o1 (70B) | 61.1% | Official leaderboard | ||
| 12 | | 60.5% | Jun 2025 | Official leaderboard | |
| 13 | | 52.4% | Jul 2024 | Official leaderboard | |
| 14 | MedGemma (27B) | 51.3% | Official leaderboard | ||
| 15 | HuatuoGPT-o1 (8B) | 45.5% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 95.3% | Feb 2026 | Official leaderboard | |
| 2 | GPT-5.5 | 94.6% | Apr 2026 | Official leaderboard | |
| 3 | Claude Opus 4.7 | 92.3% | Apr 2026 | Official leaderboard | |
| 4 | GPT-5 | 90.2% | Aug 2025 | Official leaderboard | |
| 5 | | 85.8% | Sep 2025 | Official leaderboard | |
| 6 | Claude Sonnet 4.6 | 84.5% | Feb 2026 | Official leaderboard | |
| 7 | | 83.7% | May 2025 | Official leaderboard | |
| 8 | gpt-oss-120b | 79% | Aug 2025 | Official leaderboard | |
| 9 | | 76.7% | Apr 2025 | Official leaderboard | |
| 10 | GPT-5 nano | 71.9% | Aug 2025 | Official leaderboard | |
| 11 | | 60.5% | Jun 2025 | Official leaderboard | |
| 12 | | 52.4% | Jul 2024 | Official leaderboard |
Rankings by part
We read 10 parts from the benchmark’s paper, each ranking models on one part of the test, such as Clause accuracy, Evidence citation F1 and Boundary hit rate. GPT-5.5 leads 5 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Clause accuracy
12 models we track have a score here.
- 1 GPT-5.5 ↑1 98.5%
- 2 GPT-5 ↑2 98%
- 3 Claude Sonnet 4.6 ↑3 97.9%
- 4
Gemini 3.1 Pro ↓3 97.8%
- 5 Claude Opus 4.7 ↓2 97.3%
- 6
Gemini 2.5 Flash Sep 2025 ↓1 96.7%
- 7
DeepSeek-R1 May 2025 95.3%
- 8 gpt-oss-120b 94.3%
- 9
Qwen3-235B-A22B 93.3%
- 10
Mistral Small 3.2 ↑1 89.5%
Evidence citation F1
12 models we track have a score here.
- 1 GPT-5.5 ↑1 74.3%
- 2 Claude Opus 4.7 ↑1 67.4%
- 3 GPT-5 ↑1 66.4%
- 4 Claude Sonnet 4.6 ↑2 65.8%
- 5
Gemini 3.1 Pro ↓4 65.5%
- 6
Gemini 2.5 Flash Sep 2025 ↓1 64.9%
- 7
Qwen3-235B-A22B ↑2 60.6%
- 8
DeepSeek-R1 May 2025 ↓1 56.3%
- 9
Mistral Small 3.2 ↑2 49.4%
- 10 gpt-oss-120b ↓2 45.4%
Boundary hit rate
12 models we track have a score here.
- 1 GPT-5.5 ↑1 88%
- 2 Claude Sonnet 4.6 ↑4 85.8%
- 3 GPT-5 ↑1 83.9%
- 4
Qwen3-235B-A22B ↑5 81.6%
- 5
Gemini 3.1 Pro ↓4 78.6%
- 6 Claude Opus 4.7 ↓3 78.5%
- 7
Gemini 2.5 Flash Sep 2025 ↓2 77.5%
- 8
DeepSeek-R1 May 2025 ↓1 76.5%
- 9 GPT-5 nano ↑1 72.5%
- 10
Mistral Small 3.2 ↑1 71.4%
Missing detection F1
12 models we track have a score here.
- 1
Gemini 3.1 Pro 88.2%
- 2 Claude Opus 4.7 ↑1 83.4%
- 3 GPT-5.5 ↓1 82.2%
- 4 GPT-5 77.7%
- 5
Gemini 2.5 Flash Sep 2025 70.5%
- 6 Claude Sonnet 4.6 70.1%
- 7
DeepSeek-R1 May 2025 66.5%
- 8
Qwen3-235B-A22B ↑1 54.7%
- 9
Llama 3.1-8B ↑3 46.7%
- 10 gpt-oss-120b ↓2 36.5%
Missing slot identification F1
12 models we track have a score here.
- 1 Claude Opus 4.7 ↑2 77.6%
- 2
Gemini 3.1 Pro ↓1 77.4%
- 3 GPT-5.5 ↓1 77.3%
- 4 Claude Sonnet 4.6 ↑2 76.2%
- 5 GPT-5 ↓1 75%
- 6 gpt-oss-120b ↑2 71.6%
- 7
Gemini 2.5 Flash Sep 2025 ↓2 71.6%
- 8
Qwen3-235B-A22B ↑1 69.3%
- 9
DeepSeek-R1 May 2025 ↓2 66.3%
- 10 GPT-5 nano 61%
Uncertain detection F1
12 models we track have a score here.
- 1 GPT-5.5 ↑1 79.7%
- 2
Gemini 3.1 Pro ↓1 74%
- 3 Claude Opus 4.7 57.1%
- 4 Claude Sonnet 4.6 ↑2 49.2%
- 5
DeepSeek-R1 May 2025 ↑2 36.3%
- 6
Gemini 2.5 Flash Sep 2025 ↓1 31.3%
- 7
Qwen3-235B-A22B ↑2 18.3%
- 8 GPT-5 ↓4 12.8%
- 9
Llama 3.1-8B ↑3 10.5%
- 10 GPT-5 nano 10.1%
Reportable detection F1
12 models we track have a score here.
- 1
Gemini 3.1 Pro 94%
- 2 GPT-5.5 91.7%
- 3 Claude Opus 4.7 90.9%
- 4 GPT-5 86.9%
- 5
Gemini 2.5 Flash Sep 2025 82.4%
- 6
DeepSeek-R1 May 2025 ↑1 82%
- 7 Claude Sonnet 4.6 ↓1 81.6%
- 8
Qwen3-235B-A22B ↑1 74.4%
- 9 gpt-oss-120b ↓1 71.4%
- 10 GPT-5 nano 69.4%
Complete
12 models we track have a score here.
- 1
Gemini 3.1 Pro 98.6%
- 2 GPT-5.5 97.8%
- 3 GPT-5 ↑1 97%
- 4 Claude Opus 4.7 ↓1 95.8%
- 5
Gemini 2.5 Flash Sep 2025 92.7%
- 6
DeepSeek-R1 May 2025 ↑1 92.4%
- 7 gpt-oss-120b ↑1 88.8%
- 8 Claude Sonnet 4.6 ↓2 88.3%
- 9
Qwen3-235B-A22B 82.6%
- 10 GPT-5 nano 82%
Missing
12 models we track have a score here.
- 1
Gemini 3.1 Pro 92.4%
- 2 GPT-5.5 91.9%
- 3 Claude Opus 4.7 89.9%
- 4 GPT-5 88.7%
- 5 Claude Sonnet 4.6 ↑1 82%
- 6
Gemini 2.5 Flash Sep 2025 ↓1 79.2%
- 7
Qwen3-235B-A22B ↑2 72.2%
- 8
DeepSeek-R1 May 2025 ↓1 70%
- 9 gpt-oss-120b ↓1 67.9%
- 10 GPT-5 nano 57.3%
Uncertain
12 models we track have a score here.
- 1 GPT-5.5 ↑1 66.5%
- 2
Gemini 3.1 Pro ↓1 65.4%
- 3 Claude Opus 4.7 56.8%
- 4 Claude Sonnet 4.6 ↑2 46.3%
- 5
DeepSeek-R1 May 2025 ↑2 37.7%
- 6
Llama 3.1-8B ↑6 29.2%
- 7
Gemini 2.5 Flash Sep 2025 ↓2 26.8%
- 8
Qwen3-235B-A22B ↑1 21%
- 9 GPT-5 nano ↑1 12.5%
- 10 GPT-5 ↓6 7.80%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The PSEBench paper
GPT-5 on the official leaderboard passed 85% in August 2025
Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official PSEBench 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
7 of 8 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the benchmark’s paper (15 entries).
- Pass Models scored 12 models we track, plus 3 other entries on its leaderboard.
- Note Newest score The newest came 7 months ago, in April 2026, for GPT-5.5. That’s expected for a saturated test.
- Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-5.5 (#21).
- Note Headroom Saturated. The best is 95.3%, past 85% since August 2025.
Leaderboard import 3 pass
- Pass Importer We read the board on labs.scale.com: 12 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.