Medical AI is moving past generic chatbots to deliver specialized search tools that clinicians actually trust.

The battle for the clinical desktop is no longer about general knowledge. It is about speed, depth, and safety. OpenEvidence has released a new family of medical AI search models—Osler, Sackett, and Snow—designed to match different clinical paces.

Some tasks need a five-second answer at the bedside. Others require an exhaustive literature review. By tiering these models, the technology adapts to the doctor, not the other way around. This pragmatism is what clinical AI has desperately lacked.

The Power of Gatekeeping

The company also developed Darwin, a model that scored a perfect 100% on a medically reviewed MedQA benchmark. Yet, you cannot easily use it.

Because of safety concerns regarding dual-use capabilities, Darwin is locked behind an application-only gate. This decision highlights a growing tension in healthcare AI. The most powerful tools are now deemed too risky for open access. It signals a shift from open-source optimism to strict risk mitigation.

Earning Clinical Trust

Doctors are notoriously skeptical of AI, and for good reason. Hallucinations in medicine can be fatal. However, data from the Stanford-Harvard NOHARM study showed clinicians preferred this platform over all other external AI chatbots combined.

A new partnership with a leading cancer center to integrate precision oncology will test this trust in high-stakes environments. If AI can accurately parse genomic data at the point of care, it moves from a search tool to an essential clinical partner. The real test is whether restricted access models like Darwin can scale without compromising safety.