Orinn consistently outperforms leading state-of-the-art models across critical medical benchmarks, setting a new standard for clinical AI accuracy and reliability. To ensure rigorous, unbiased evaluation, each benchmark was assessed using a multi-judge framework — independent AI models scored Orinn's outputs against validated medical datasets at a controlled temperature of 0.2 for maximum consistency. Scores were then cross-validated with a secondary judge model and averaged to eliminate single-model bias. This robust evaluation pipeline reflects our commitment to building transparent, evidence-driven AI that healthcare systems can trust — today and into the future.

Orinn 1.6 demonstrates state-of-the-art performance across key healthcare AI benchmarks, validating its strength in real-world clinical intelligence. In medical knowledge and reasoning, it achieves leading scores on MedQA (95.8) and PubMedQA (95.3), reflecting deep clinical understanding and diagnostic accuracy. For document and long-context processing, Orinn delivers top-tier results on OmniDocBench 1.5 (88.5) and OCRBench (89.5), enabling reliable interpretation of complex medical records and reports. In medical visual question answering, strong performance on SLAKE (85.5) highlights its ability to understand multimodal clinical data. Additionally, Orinn maintains low hallucination rates and high consistency across HealthBench evaluations, ensuring trustworthy outputs. Together, these results position Orinn as a highly reliable, benchmark-validated medical AI system built for deployment across real healthcare environments.