Introducing the OpenEvidence Model Family

September 3, 2026

OpenEvidence Darwin is our most powerful model yet and the best medical AI model in the world, leading on MedXpertQA (72.8%), HealthBench Professional (82.7%), and NOHARM (87.2%), and the first AI in history to score a perfect 100% on MedQA, the leading independent benchmark of medical AI.

September 3, 2026

Today, OpenEvidence is releasing a new family of medical AI models, named for founders of modern medicine: Osler, our fastest model (~5s to answer), built for the pace of point of care; Sackett, a deeper search model (~30s to answer) for questions that turn on the weight of the evidence; and Snow, our deepest production model (~5m to answer), which runs a full investigation of world medical literature before it writes a word.

Every model in the family is held to the same standard of clinical accuracy. What varies is time: how long a model thinks, and how deep it searches.

The names are commitments.

William Osler moved medicine from the lecture hall to the bedside. OpenEvidence Osler is the model at the bedside, answering at the speed of the patient encounter. Osler is the successor to the model that currently powers OpenEvidence answers, and will continue to be the default model on OpenEvidence.

David Sackett is remembered as the father of evidence-based medicine. OpenEvidence Sackett is the model that reads the way Sackett taught: interactively asking the clinician for further clinical context and patiently and methodically weighing all of the evidence.

John Snow founded modern epidemiology when he traced the 1854 Broad Street cholera outbreak, case by case, to a single water pump. OpenEvidence Snow, the successor to OpenEvidence Deep Consult, operates as a team of equally tenacious Ph.D.-level experts in your pocket. OpenEvidence Snow meticulously runs down every line of inquiry through the literature, pursuing multiple paths of research in parallel, before it commits to a report.

Osler, Sackett, and Snow are rolling out to all OpenEvidence users today—free to verified clinicians. All three models are available on openevidence.com and the OpenEvidence iOS/Android Apps, allowing clinicians to dynamically select and match the model to the time they have available. There is a fourth model.

OpenEvidence Darwin is the most advanced medical AI model in the world, now in research preview. It is the first AI model in history to achieve a perfect score on MedQA—the leading fully-independent medical AI benchmark. We are publishing the results alongside Darwin's explanation for every answer. Darwin is also state-of-the-art on every other leading benchmark of medical AI: MedXpertQA (72.8%), HealthBench Professional (82.7%), and NOHARM (87.2%), ahead of the next-best models (Claude Fable 5 and Gemini 3.7).

Darwin is currently in research preview. Capability in medicine is inherently dual-use: a model that can reason at the frontier of virology, immunology, and human genetics could also, in the wrong hands, accelerate work the world tightly governs for good reason, such as bioweapons-relevant research and germline editing outside mainstream scientific oversight. Consistent with the safety frameworks now standard at the AI frontier, Darwin is available by application only. Today that means institutional partners such as the National Organization for Rare Disorders (NORD), which brings Darwin its hardest cases; research collaborators; and accredited AI researchers at academic institutions benchmarking the accuracy and safety of AI in clinical medicine. As Darwin's safeguards are validated with these partners, its capabilities will flow, model by model, into Osler, Sackett, and Snow — the frontier arriving at the point of care as fast as it can arrive safely.

The OpenEvidence Model Family represents the next generation of medical AI search models, available today for free for verified clinicians.

← See all announcements