Llama3-Med42-70B: benchmark results
M42 Health's clinical fine-tune of Llama 3 70B for medical QA and clinical decision support, released for research rather than clinical use. Provider: Meta. Released 2024-06-27. Access: Open.
Aggregate Rating 1554 ± 27, rank #605 of 1654 rated models, from 25 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - BBH | 66.88 | Score | 95.1 |
| Open LLM Leaderboard - MuSR | 46.29 | Score | 88.9 |
| Open LLM Leaderboard - MMLU-Pro | 49.63 | Score | 87.5 |
| Open LLM Leaderboard - GPQA | 34.73 | Score | 87.1 |
| Open LLM Leaderboard - MATH Level 5 | 22.58 | Score | 75.9 |
| Open LLM Leaderboard - IFEval | 62.91 | Score | 75 |
| MedPriv-Bench (Privacy Prompt) - Response Utility | 4.6 | Response utility (0-5): GPT-5-mini grades each answer agains | 64.3 |
| EHRStruct (Synthea) - Count | 18 | Accuracy (%; D-R1: 100 zero-shot questions on synthetic Synt | 50 |
| EHRStruct (Synthea) - Filtering (Greater Than) | 13 | Accuracy (%; D-U1: 100 zero-shot questions on synthetic Synt | 50 |
| MedPriv-Bench - Response Utility | 4.54 | Response utility (0-5): GPT-5-mini grades each answer agains | 50 |
| EHRStruct (Synthea) - Subtraction | 27 | Accuracy (%; D-R4: 100 zero-shot questions on synthetic Synt | 46.7 |
| MEDIC Benchmark | 61.34 | MEDIC Public Table Average | 44.2 |
Interactive version: theaggregate.ai/model?slug=llama3-med42-70b · How It Works · Data refreshed daily, snapshot 2026-10-24.