HELM MedHELM - PubMedQA: leaderboard
HELM MedHELM run of PubMedQA, yes/no/maybe question answering over biomedical research abstracts.
Metric: EM. Source: crfm.stanford.edu. Saturation forecast: Around January 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 05-06) | 76.4 | #147 |
| 2 | GPT-5 | 74.8 | #98 |
| 3 | Claude 3.5 Sonnet (20241022) | 74.4 | #291 |
| 4 | DeepSeek R1 | 74.3 | #251 |
| 5 | O3 Mini (2025-01-31) | 74.3 | #240 |
| 6 | O4 Mini (2025-04-16) | 74 | #179 |
| 7 | Llama 3.3 70B Instruct | 73.3 | #534 |
| 8 | GPT-5 Mini (2025-08-07) | 72 | #172 |
| 9 | GPT-4o (2024-05-13) | 70.1 | #325 |
| 10 | GPT-4o Mini (2024-07-18) | 67.2 | #558 |
| 11 | Gemini 2.0 Flash | 64 | #342 |
| 12 | Claude 3.7 Sonnet (20250219) | 62.5 | #201 |
| 13 | Gemini 1.5 Pro (001) | 62.5 | #444 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-medhelm-pubmedqa · How It Works · Data refreshed daily, snapshot 2026-10-20.