A team led by Andrew Zhang and colleagues announced Clinical System One (CS1-4B), a decision model for health records. It has 4 billion parameters and was trained on 12.4 million question‑answer pairs from 3.6 million patients. The model can answer multiple typed questions in a single forward pass and provides a probability for each answer, letting clinicians review uncertain cases. In tests on 963,408 held‑out patients, CS1‑4B beat larger open models on 13 of 14 tasks and was 6.4 times faster on external data sets.
Why it matters
Hospitals could use CS1‑4B to screen patients for trial eligibility faster and with fewer errors, freeing clinicians to focus on care.
