Benchmark Almanac

MedSafe-Dx v0.3

Claude Opus 5.5 73.5 leads since October 2026
Not in model cards

Models read synthetic primary-care intake summaries and decide whether to flag a dangerous condition. Rule-based labels determine weighted error costs, with missed serious cases costing seven times a false escalation or incorrect danger label. The score normalizes these costs against a fixed escalation baseline.

Version 0.3 changes case labels and is not comparable with the original preprint. The current board uses 900 synthetic cases and 20 models. Several evaluated models helped draft or validate the rules. Clinical advisor review remains in progress.

Released
October 2026
Kind
Medicine
Level
Expert
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Generated
Run by
Anyone
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
September 2026
Scores refreshed
November 19th, 2026
20
Entries on its own leaderboard, 19 models we track

Claude Opus 5.5 leads with 73.5 among models released by September 2026 Claude Opus 5.5 leads with 73.5 among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 19th, 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 September 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 Claude Opus 5.5 v0.3 full, prompt 4aj 73.5
  2. 2 Claude Fable 5.1 v0.3 full, prompt 4aj 71.1
  3. 3 GPT-6.1 Sol v0.3 full, prompt 4aj 69.2
  4. 4 GPT-6 Astra v0.3 full, prompt 4aj 68.8
  5. 5 Gemini 3.1 Pro v0.3 full, prompt 4aj 68.5
  6. 6 GPT-6 Luna v0.3 full, prompt 4aj 67.1
  7. 7 GPT-5.6 Terra v0.3 full, prompt 4aj 64.2
  8. 8 Gemini 3.8 Flash v0.3 full, prompt 4aj 61.5
  9. 9 GPT-5.2 v0.3 full, prompt 4aj 60.9
  10. 10 Grok 4.7 v0.3 full, prompt 4aj 60.8
  11. 11 Kimi K3 v0.3 full, prompt 4aj 58.7
  12. 12 GPT-5.4 Mini v0.3 full, prompt 4aj 57.4
  13. 13 Claude Sonnet 5.5 v0.3 full, prompt 4aj 57.2
  14. 14 DeepSeek V4.1 Flash v0.3 full, prompt 4aj 55.1
  15. 15 Claude Sonnet 4.6 v0.3 full, prompt 4aj 50.6
  16. 16 GLM-5.3 v0.3 full, prompt 4aj 40.5
  17. 17 gpt-oss-120b v0.3 full, prompt 4aj 33.6
  18. 18 Claude Haiku 4.5 v0.3 full, prompt 4aj 19.9
  19. 19 Llama 4 Maverick v0.3 full, prompt 4aj 11.9
  20. 20 llama-3.1-8b-instruct v0.3 full, prompt 4aj -116.60

Source: Official MedSafe-Dx v0.3 leaderboard

#EntryScore ± 95% CISettingDateSource
1 Claude Opus 5.5 73.5 69.5 to 77.4 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
2 Claude Fable 5.1 71.1 66 to 75.9 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
3 GPT-6.1 Sol 69.2 ±3.9 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
4 GPT-6 Astra 68.8 ±4.1 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
5 Gemini 3.1 Pro 68.5 63.6 to 73.2 v0.3 full, prompt 4aj Feb 2026 Official leaderboard
6 GPT-6 Luna 67.1 62.2 to 71.6 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
7 GPT-5.6 Terra 64.2 58.5 to 69.4 v0.3 full, prompt 4aj Jul 2026 Official leaderboard
8 Gemini 3.8 Flash 61.5 56.2 to 66.5 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
9 GPT-5.2 60.9 55.8 to 65.6 v0.3 full, prompt 4aj Dec 2025 Official leaderboard
10 Grok 4.7 60.8 55.2 to 65.9 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
11 Kimi K3 58.7 53.1 to 64.5 v0.3 full, prompt 4aj Jul 2026 Official leaderboard
12 GPT-5.4 Mini 57.4 ±6 v0.3 full, prompt 4aj Mar 2026 Official leaderboard
13 Claude Sonnet 5.5 57.2 51.8 to 62.4 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
14 DeepSeek V4.1 Flash 55.1 48.7 to 61 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
15 Claude Sonnet 4.6 50.6 43.2 to 57.5 v0.3 full, prompt 4aj Feb 2026 Official leaderboard
16 GLM-5.3 40.5 32.7 to 47.9 v0.3 full, prompt 4aj Aug 2026 Official leaderboard
17 gpt-oss-120b 33.6 ±7.1 v0.3 full, prompt 4aj Aug 2025 Official leaderboard
18 Claude Haiku 4.5 19.9 10.9 to 29.4 v0.3 full, prompt 4aj Oct 2025 Official leaderboard
19 Llama 4 Maverick 11.9 ±9.1 v0.3 full, prompt 4aj Apr 2025 Official leaderboard
20 llama-3.1-8b-instruct -116.60 -127.9 to -104.4 v0.3 full, prompt 4aj Official leaderboard
#ModelScore ± 95% CISettingDateSource
1 Claude Opus 5.5 73.5 69.5 to 77.4 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
2 Claude Fable 5.1 71.1 66 to 75.9 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
3 GPT-6.1 Sol 69.2 ±3.9 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
4 GPT-6 Astra 68.8 ±4.1 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
5 Gemini 3.1 Pro 68.5 63.6 to 73.2 v0.3 full, prompt 4aj Feb 2026 Official leaderboard
6 GPT-6 Luna 67.1 62.2 to 71.6 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
7 GPT-5.6 Terra 64.2 58.5 to 69.4 v0.3 full, prompt 4aj Jul 2026 Official leaderboard
8 Gemini 3.8 Flash 61.5 56.2 to 66.5 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
9 GPT-5.2 60.9 55.8 to 65.6 v0.3 full, prompt 4aj Dec 2025 Official leaderboard
10 Grok 4.7 60.8 55.2 to 65.9 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
11 Kimi K3 58.7 53.1 to 64.5 v0.3 full, prompt 4aj Jul 2026 Official leaderboard
12 GPT-5.4 Mini 57.4 ±6 v0.3 full, prompt 4aj Mar 2026 Official leaderboard
13 Claude Sonnet 5.5 57.2 51.8 to 62.4 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
14 DeepSeek V4.1 Flash 55.1 48.7 to 61 v0.3 full, prompt 4aj Sep 2026 Official leaderboard
15 Claude Sonnet 4.6 50.6 43.2 to 57.5 v0.3 full, prompt 4aj Feb 2026 Official leaderboard
16 GLM-5.3 40.5 32.7 to 47.9 v0.3 full, prompt 4aj Aug 2026 Official leaderboard
17 gpt-oss-120b 33.6 ±7.1 v0.3 full, prompt 4aj Aug 2025 Official leaderboard
18 Claude Haiku 4.5 19.9 10.9 to 29.4 v0.3 full, prompt 4aj Oct 2025 Official leaderboard
19 Llama 4 Maverick 11.9 ±9.1 v0.3 full, prompt 4aj Apr 2025 Official leaderboard

Number one over time

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

Anthropic
Own leaderboard Not tested since Oct 2026

Source: Official MedSafe-Dx v0.3 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
0.00 20.0 40.0 60.0 80.0 2025 2026 0.00 20.0 40.0 60.0 80.0 2025 2026

Source: Official MedSafe-Dx v0.3 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

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (20 entries).
  • Pass Models scored 19 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from September 2026, for GPT-6.1 Sol.
  • Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 19 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 19th, 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