Latest / Health and life sciences

Model releaseHealthFrontierUSConfirmed

Google says Med-PaLM 2 scores 85.4% on US medical licensing-style questions

See how this story connects

Google presented Med-PaLM 2, its medical large language model, at its The Check Up event on 14 March 2023. Google said the model scored 85.4% on questions in the style of the US Medical Licensing Exam, up from the 67.6% it gave for the first Med-PaLM, and called the result expert level. Clinician reviews found its answers were sometimes less complete than doctors’ answers.

Why it matters

Health systems and regulators got a public exam benchmark for a medical language model from a large cloud vendor, alongside clinician reviews showing where its answers fell short.

When
14 Mar 2023
ID
WTA-N-2023-0047
Who
Google, Google Research

Summary by wheresthe.ai, from Google Research and the reporting credited above. Checked against the source on 9 Oct 2026.

Source copy: download the page as saved on 18 Oct 2026 (PDF). Also held by the Internet Archive.

Next stop on this Storyline: What happened nextNature publishes Google’s Med-PaLM paper on how language models answer medical questions

Cite this story

WTA-N-2023-0047: “Google says Med-PaLM 2 scores 85.4% on US medical licensing-style questions.” wheresthe.ai, 14 Mar 2023. https://wheresthe.ai/d/google-med-palm-2-announced-2023-03/ (source: Google Research).

Spotted an error? Email [email protected].