Benchmark Almanac

MedHallu

GPT-4o (snapshot unspecified) 0.88 tops its leaderboard Gemma 2 2B IT 0.71 leads the models we track
Not in model cards

Models distinguish factual medical answers from hallucinated answers at three difficulty levels. The benchmark tests detection with and without supporting knowledge.

The February 2025 paper compares GPT, Qwen, Gemma, Llama and medical models. GPT-4o led on hard examples without supporting knowledge at 0.625 F1.

Released
February 2025
Kind
Hallucination rates
Level
University
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
Last model added
July 2024
Scores refreshed
November 20th, 2026
14
Entries on its own leaderboard, 1 model we track

GPT-4o (snapshot unspecified) tops the leaderboard with 0.88 Gemma 2 2B IT leads with 0.71

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. The best model we track is Gemma 2 2B IT, at 0.71. This view shows only models we track, each at its best.

  1. 1 GPT-4o (snapshot unspecified) With Knowledge 0.88
  2. 2 Qwen2.5-14B-Instruct With Knowledge 0.85
  3. 3 GPT-4o mini (snapshot unspecified) With Knowledge 0.84
  4. 4 Qwen2.5-7B-Instruct With Knowledge 0.84
  5. 5 Gemma-2-9b-Instruct With Knowledge 0.84
  6. 6 DeepSeek-R1-Distill-Llama-8B With Knowledge 0.81
  7. 7 Llama-3.1-8B-Instruct With Knowledge 0.80
  8. 8 Llama3-Med42-8B With Knowledge 0.80
  9. 9 Llama-3.1-8B-UltraMedical With Knowledge 0.77
  10. 10 Llama-3.2-3B-Instruct With Knowledge 0.73
  11. 11 Gemma 2 2B IT With Knowledge 0.71
  12. 12 Qwen2.5-3B-Instruct With Knowledge 0.68
  13. 13 BioMistral-7B With Knowledge 0.65
  14. 14 OpenBioLLM-Llama3-8B Without Knowledge 0.48

Source: Official MedHallu leaderboard

Source: Official MedHallu leaderboard

#EntryScoreSettingDateSource
1 GPT-4o (snapshot unspecified) 0.88 With Knowledge Official leaderboard
2 Qwen2.5-14B-Instruct 0.85 With Knowledge Official leaderboard
3 GPT-4o mini (snapshot unspecified) 0.84 With Knowledge Official leaderboard
4 Qwen2.5-7B-Instruct 0.84 With Knowledge Official leaderboard
5 Gemma-2-9b-Instruct 0.84 With Knowledge Official leaderboard
6 DeepSeek-R1-Distill-Llama-8B 0.81 With Knowledge Official leaderboard
7 Llama-3.1-8B-Instruct 0.80 With Knowledge Official leaderboard
8 Llama3-Med42-8B 0.80 With Knowledge Official leaderboard
9 Llama-3.1-8B-UltraMedical 0.77 With Knowledge Official leaderboard
10 Llama-3.2-3B-Instruct 0.73 With Knowledge Official leaderboard
11 Gemma 2 2B IT 0.71 With Knowledge Jul 2024 Official leaderboard
12 Qwen2.5-3B-Instruct 0.68 With Knowledge Official leaderboard
13 BioMistral-7B 0.65 With Knowledge Official leaderboard
14 OpenBioLLM-Llama3-8B 0.48 Without Knowledge Official leaderboard
#ModelScoreSettingDateSource
1 Gemma 2 2B IT 0.71 With Knowledge Jul 2024 Official leaderboard

Scores over time

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

On the benchmark’s own leaderboard
0.55 0.60 0.65 0.70 0.75 0.55 0.60 0.65 0.70 0.75

Source: Official MedHallu 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 Factuality

All 164 in Factuality
  • SimpleVQA In use Models answer 2,025 short factual questions about images across 9 tasks and 9 topics. An LLM judge grades answers against fixed references. Qwen3.7-Plus 81.7%
  • HallusionBench In use Models answer 1,129 yes or no questions about 346 images, built in pairs to expose language hallucination and visual illusion. Scored by accuracy. Keye-VL-2.0-30B-A3B 73.5%
  • SimpleQA Verified In use 1,000 human-verified, de-duplicated short-answer factual questions measuring parametric knowledge. Gemini 3.1 Pro 77.5%
  • AA-Omniscience v1.1 In use Models answer 3,000 knowledge questions in six domains without tools and may abstain. The index adds correct answers and subtracts wrong ones. Inkling 2 Small 16.1
  • GaRAGe In use Models write long-form RAG answers to 2,366 questions with human-annotated grounding passages. Scoring checks whether answers use the relevant passages. Amazon Nova Pro 60.7%
  • Kimi Hallu Bench In use Models answer hard prompts drawn from practical work tasks. Fact-checking agents score the share of claims that are hallucinated. GPT-6 Astra 4.20%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in July 2024, for Gemma 2 2B IT. 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 (14 entries).
  • Pass Models scored 1 model we track, plus 13 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in July 2024, for Gemma 2 2B IT.
  • Note Frontier models None of its models are on the capabilities index.
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on medhallu.github.io: 2 rows for models we track and 13 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