HELM MedHELM - EHRSHOT: leaderboard
HELM MedHELM run of EHRSHOT, few-shot clinical prediction over structured electronic health-record timelines.
Metric: EM. Source: crfm.stanford.edu. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o Mini (2024-07-18) | 90.67 | #558 |
| 2 | Llama 3.3 70B Instruct | 64.43 | #534 |
| 3 | O3 Mini (2025-01-31) | 64.2 | #240 |
| 4 | O4 Mini (2025-04-16) | 54.17 | #179 |
| 5 | GPT-4o (2024-05-13) | 54.13 | #325 |
| 6 | GPT-5 Mini (2025-08-07) | 51.1 | #172 |
| 7 | Gemini 2.0 Flash | 50.27 | #342 |
| 8 | Gemini 1.5 Pro (001) | 50 | #444 |
| 9 | Claude 3.7 Sonnet (20250219) | 44.97 | #201 |
| 10 | GPT-5 | 36.73 | #98 |
| 11 | Claude 3.5 Sonnet (20241022) | 35.83 | #291 |
| 12 | DeepSeek R1 | 22.03 | #251 |
| 13 | Gemini 2.5 Pro (Preview 05-06) | 18.47 | #147 |
Interactive version: theaggregate.ai/benchmark?slug=helm-medhelm-ehrshot · How It Works · Data refreshed daily, snapshot 2026-10-20.