OpenEvidence Expands Oncology AI With Precision Knowledge Base, Launches New Model Family

The partnership will add expert-curated interpretations of cancer genomic alterations, patient-specific context and peer-reviewed evidence to OpenEvidence’s oncology workflows.

OpenEvidence is expanding its oncology capabilities by integrating a precision oncology knowledge base from a leading US cancer center into its clinical AI platform, while launching a new family of medical AI models for clinicians.

The partnership will add expert-curated interpretations of cancer genomic alterations, patient-specific context and peer-reviewed evidence to OpenEvidence’s oncology workflows. The company has not yet named the cancer center and said it plans to identify the organization next week.

The move builds on OpenEvidence’s existing oncology capabilities, including its licensing agreement with the National Comprehensive Cancer Network (NCCN) and integration of ASCO guidelines, figures and flowcharts. The new knowledge base will be incorporated into OpenEvidence’s specialized oncology AI sub-agent, which the company said has digitized NCCN treatment algorithms.

OpenEvidence founder Daniel Nadler said the company is working with several leading cancer centers, with additional partnerships expected to be announced in the coming weeks. The company’s longer-term strategy is to develop specialized AI agents across medical subspecialties, beginning with oncology and expanding into genetics, cardiology and neurology.

The company also launched a new family of medical AI models, led by OpenEvidence Darwin, which is currently available in research preview. OpenEvidence said Darwin achieved a 100% score on MedQA, along with scores of 72.8% on MedXpertQA, 82.7% on HealthBench Professional and 87.2% on NOHARM. The company said Darwin is currently available by application to institutional partners, research collaborators and accredited AI researchers at academic institutions.

For point-of-care use, OpenEvidence is rolling out three production models: Osler, Sackett and Snow. Osler is designed for faster responses and will remain the platform’s default model. Sackett is designed for deeper evidence searches, while Snow conducts broader investigations of medical literature before producing a report.

OpenEvidence said the three models are available to all verified clinicians at no cost. As of September, the platform had 1.12 million US licensed and verified clinicians using its services and was tracking more than 40 million NPI-verified queries from US clinicians over a 30-day period.

The company has also expanded beyond clinical search into documentation, medical coding, voice-based clinical decision support and physician communication workflows.

Stay tuned for more such updates on Digital Health News