MAI-DxO (Diagnostic Orchestrator)
Orchestrator where LLM roles act as a virtual physician panel; with o3 it solves 85.5% of 304 NEJM cases vs 20% for 21 physicians.
Frames medical diagnosis as sequential ordering of tests and questions under a cost budget, claiming higher accuracy and lower testing cost than physicians or any single model. Research demo only, not cleared for clinical use.
- Date
- Monday, 30 June 2025
- Lab
- Microsoft
- Kind
- product
- Access
- research preview
Figures
| Measure | Value | Measured by |
|---|---|---|
| NEJM case accuracy (MAI-DxO + o3) | 85.5% of 304 cases physician mean 20%, physicians barred from colleagues, books and AI | company |
Company-run study; physician comparison setup is restrictive.
Sources
This record was checked against its sources on 6 October 2026. How we check