Quadrant Health AI becomes the highest scoring AI in history on the United States Medical Licensing Examination (USMLE)

Quadrant Health AI, the medically-tuned generative artificial intelligence powering the patient-facing conversational AI, call center AI, and scribe AI at Quadrant Health, has achieved the highest score of any AI in history on the United States Medical Licensing Examination (USMLE) - 93.23%.

Across all the 3 USMLE Exams, Quadrant Health AI made 27.2% fewer errors than OpenEvidence, 83.6% fewer errors than ChatGPT, 44.7% fewer errors than GPT-4, 79.8% fewer errors than Claude 2, and 49.9% fewer errors than MedPaLM 2.

This achievement demonstrates the enormous clinical knowledge that Quadrant Health AI possesses and its potential to transform patient care and the broader healthcare industry.

Above, we show the performance of Quadrant Health AI per Step exam, relative to OpenEvidence, the previous record holder for USMLE performance.

Methodology

To evaluate our model's proficiency on the USMLE exams, we employ the official sample examination accessible at usmle.org. We do not factor in questions featuring images. This method aligns with the well-established precedent set by other AI's that have been evaluated on USMLE, as we adhere to the same question set as they have evaluated their AI's on. Thus, our evaluation encompasses 94 questions from Step 1, 109 from Step 2, and 122 from Step 3, totaling 325 questions.

Takeaways

Quadrant Health is powered by clinical AI that is best-in-class. Supercharge your practice and provide your patients with evidence-based care using our platform.

For more information about the solutions that Quadrant Health provides, click here.

To request a demo or get in touch with our team, click here.