MediPhi: benchmark results
Microsoft Healthcare and Life Sciences' MediPhi medical and clinical research model, merging five Phi-3.5-mini-instruct-derived domain experts with BreadCrumbs. This unified expert is distinct from the subsequently clinically aligned MediPhi-Instruct release. Provider: Microsoft. Access: Open.
Aggregate Rating 1355 ± 49, rank #7572 of 10916 rated models, from 10 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - ASSIN2 RTE | 91.87 | Macro F1 (%) | 71.4 |
| Open Portuguese LLM - ENEM | 66.9 | Accuracy (%) | 65 |
| Open Portuguese LLM - BLUEX | 55.77 | Accuracy (%) | 64.1 |
| Open Portuguese LLM - OAB Exams | 46.01 | Accuracy (%) | 62.7 |
| Open Portuguese LLM - FaQuAD NLI | 55.37 | Macro F1 (%) | 37.6 |
| MedLoCoMo | 9.2 | Score (%) over all 17,892 questions: item-weighted combinati | 8.3 |
| MedLoCoMo - Answerable Accuracy | 13.9 | LLM-judge accuracy (%) on the answerable questions of both s | 8.3 |
| MedLoCoMo - Cross-Admission | 8.1 | Score (%) on the 8,946 cross-admission questions (evidence f | 8.3 |
| MedLoCoMo - Single-Admission | 10.4 | Score (%) on the 8,946 single-admission questions: item-weig | 8.3 |
| MEDIC Benchmark | 32.03 | MEDIC Public Table Average | 3.2 |
Interactive version: theaggregate.ai/model?slug=mediphi · How It Works · Data refreshed daily, snapshot 2026-11-08.