Benchmark Almanac

Medbullets

GPT-4 Jun 2023 68.8% leads since February 2024
Not in model cards

Models answer clinical questions patterned on US medical licensing examinations. The benchmark also evaluates explanations of the answers.

Made by
Hanjie Chen and colleagues
Released
February 2024
Kind
Medicine
Level
Expert
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Llama 3-70B #182 on the capability index
Last model added
April 2024
Scores refreshed
November 20th, 2026
7
Entries in its paper, 4 models we track

GPT-4 leads with 68.8% among models released by April 2024 GPT-4 leads with 68.8% among models released by April 2024

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. 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 April 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-4 Jun 2023, 4 options, zero-shot explain then answer 68.8%
  2. 2 Llama 3-70B 4 options, zero-shot direct answer 68.5%
  3. 3 meerkat-7b-v1.0 4 options, zero-shot explain then answer 56.8%
  4. 4 GPT-3.5 Turbo Jun 2023, 4 options, zero-shot explain then answer 51%
  5. 5 medalpaca-13b 4 options, zero-shot direct answer 48.4%
  6. 6 chat-bison-001 4 options, zero-shot direct answer 47.7%
  7. 7 Llama 2-70B 4 options, zero-shot explain then answer 35.1%

Source: The Medbullets paper

#EntryScoreSettingDateSource
1 GPT-4 Jun 2023 68.8% 4 options, zero-shot explain then answer Jun 2023 Official leaderboard
2 Llama 3-70B 68.5% 4 options, zero-shot direct answer Apr 2024 Official leaderboard
3 meerkat-7b-v1.0 56.8% 4 options, zero-shot explain then answer Official leaderboard
4 GPT-3.5 Turbo Jun 2023 51% 4 options, zero-shot explain then answer Jun 2023 Official leaderboard
5 medalpaca-13b 48.4% 4 options, zero-shot direct answer Official leaderboard
6 chat-bison-001 47.7% 4 options, zero-shot direct answer Official leaderboard
7 Llama 2-70B 35.1% 4 options, zero-shot explain then answer Jul 2023 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-4 Jun 2023 68.8% 4 options, zero-shot explain then answer Jun 2023 Official leaderboard
2 Llama 3-70B 68.5% 4 options, zero-shot direct answer Apr 2024 Official leaderboard
3 GPT-3.5 Turbo Jun 2023 51% 4 options, zero-shot explain then answer Jun 2023 Official leaderboard
4 Llama 2-70B 35.1% 4 options, zero-shot explain then answer Jul 2023 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
GPT-4
5 options, … 4 options, …
68.8%
Llama 3-70B
5 options, … 4 options, …
68.5%
GPT-3.5 Turbo
5 options, … 5 options, … 4 options, …
51%
Llama 2-70B
5 options, … 4 options, …
35.1%

Source: The Medbullets paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-4 Jun 2023 4 options, zero-shot explain then answer 68.8%
Llama 3-70B 4 options, zero-shot direct answer 68.5%
Llama 3-70B 4 options, zero-shot explain then answer 67.2%
GPT-4 Jun 2023 4 options, zero-shot direct answer 66.2%
Llama 3-70B 5 options, zero-shot direct answer 64%
GPT-4 Jun 2023 5 options, zero-shot explain then answer 63.3%
Llama 3-70B 5 options, zero-shot explain then answer 62.7%
GPT-4 Jun 2023 5 options, zero-shot direct answer 60.7%
GPT-3.5 Turbo Jun 2023 4 options, zero-shot explain then answer 51%
GPT-3.5 Turbo Jun 2023 5 options, zero-shot explain then answer 47.4%
GPT-3.5 Turbo Jun 2023 4 options, zero-shot direct answer 46.1%
GPT-3.5 Turbo Jun 2023 5 options, zero-shot direct answer 43.2%
Llama 2-70B 4 options, zero-shot explain then answer 35.1%
Llama 2-70B 4 options, zero-shot direct answer 33.1%
Llama 2-70B 5 options, zero-shot explain then answer 27.9%
Llama 2-70B 5 options, zero-shot direct answer 26%

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
GPT-4
5 options, … 4 options, …
68.8%
Llama 3-70B
5 options, … 4 options, …
68.5%
GPT-3.5 Turbo
5 options, … 5 options, … 4 options, …
51%
Llama 2-70B
5 options, … 4 options, …
35.1%

Source: The Medbullets paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-4 Jun 2023 4 options, zero-shot explain then answer 68.8%
Llama 3-70B 4 options, zero-shot direct answer 68.5%
Llama 3-70B 4 options, zero-shot explain then answer 67.2%
GPT-4 Jun 2023 4 options, zero-shot direct answer 66.2%
Llama 3-70B 5 options, zero-shot direct answer 64%
GPT-4 Jun 2023 5 options, zero-shot explain then answer 63.3%
Llama 3-70B 5 options, zero-shot explain then answer 62.7%
GPT-4 Jun 2023 5 options, zero-shot direct answer 60.7%
GPT-3.5 Turbo Jun 2023 4 options, zero-shot explain then answer 51%
GPT-3.5 Turbo Jun 2023 5 options, zero-shot explain then answer 47.4%
GPT-3.5 Turbo Jun 2023 4 options, zero-shot direct answer 46.1%
GPT-3.5 Turbo Jun 2023 5 options, zero-shot direct answer 43.2%
Llama 2-70B 4 options, zero-shot explain then answer 35.1%
Llama 2-70B 4 options, zero-shot direct answer 33.1%
Llama 2-70B 5 options, zero-shot explain then answer 27.9%
Llama 2-70B 5 options, zero-shot direct answer 26%

Number one over time

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

OpenAI
Benchmark paper Not tested since Apr 2024
GPT-4 68.8%

Source: The Medbullets paper

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
In the benchmark’s paper
0% 20% 40% 60% 80% 100% 2023 2024 Near the ceiling (85%) GPT-4 0% 20% 40% 60% 80% 100% 2023 2024 Near the ceiling (85%) GPT-4

Source: Official Medbullets 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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in April 2024, for Llama 3-70B. 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 benchmark’s paper (7 entries).
  • Pass Models scored 4 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in April 2024, for Llama 3-70B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Llama 3-70B (#182).
  • Pass Headroom The best is 68.8%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on aclanthology.org: 16 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.

How we check