Background
We previously demonstrated that a large language model could estimate suicide risk using hospital discharge notes.
ObjectiveWith the emergence of reasoning models that can be run on consumer-grade hardware, we investigated whether these models can approximate the performance of much...