Medical Model Reads 130K-Token Patient Record in 61 Seconds, Reasons for 11 Minutes

A developer tested AntLing's Ling-3.0-flash-Sante medical model on a 71-year synthetic patient record of about 130K tokens: it read the chart in 61 seconds but took 11 minutes to reason through it.

2026-11-19 ~ 2026-11-19 · 2 related posts