Vitruviana combines Gemini 3 Pro—the first generalist AI to surpass radiology trainees—with GPT-5.1 for a hybrid architecture validated across 8 medical specialties.
Our primary model selection is backed by peer-reviewed research and verified benchmarks
On the RadLE v1 benchmark ("Radiology's Last Exam"), Gemini 3 Pro achieved 0.51 accuracy compared to 0.45 for radiology trainees—the first generalist AI to exceed trainee performance.
Med-Gemini-3D demonstrated 53% clinically acceptable CT report generation. The Med-Gemini family spans radiology, pathology, dermatology, ophthalmology, and genomics.
Gemini 3's unprecedented context window enables analysis of complete patient histories, multi-year EHR data, and comprehensive medical records in a single inference.
MedGemma provides specialized models for medical image interpretation, text comprehension, and clinical reasoning—available in multimodal and text-only variants.
Empirically validated performance from 84 real API calls across production scenarios
Intelligent routing optimizes for both reasoning depth and response speed
Primary - Clinical Reasoning
Fast - Structured Tasks
End-to-end testing of all clinical AI services
Evidence-based clinical insights with guideline citations
Drug safety analysis with polypharmacy support
Complete SOAP notes with 4/4 section compliance
Natural patient communication with clinical data capture
Real clinical scenarios tested across 8 medical specialties