OpenEvidence is letting clinicians pick how much thinking they want from a medical AI. The company has released a family of clinical search models with three production tiers plus a research preview, each named for a pioneering physician.
Osler is the default and fastest tier, replying in about five seconds. Sackett sits a level deeper, spending roughly 30 seconds on questions that hinge on the strength of the evidence. Snow replaces the older Deep Consult feature and can take five minutes to comb the literature before writing a report.
Darwin, which OpenEvidence describes as the most advanced of the family, is open to researchers by application only. The company says every tier is held to the same standard of clinical accuracy, and that the tiers differ only in how long a model thinks and how deep it searches.
The models are reaching openevidence.com and the company’s iOS and Android apps, where verified clinicians choose a tier through a model selector at no charge. The design mirrors the reasoning-model trend across AI, where more thinking time buys deeper answers and users make the trade-off per question.
