Sword Health releases Dawn 1, a mental health support model built on GLM-5.3
Overview
Sword Health releases Dawn 1, a mental health support model it built by post-training GLM-5.3 with reinforcement learning on Prime Intellect's prime-rl framework, according to Prime Intellect's announcement.
Sword says Dawn 1 delivers greater intelligence at lower inference cost than its base model, with strong performance across mental health knowledge, safety and sycophancy evaluations.
Sword's own announcement says Dawn 1 is safer and less sycophantic than the base model and outperforms frontier proprietary models on multi-turn safety protocol adherence in Sword's evaluations. These results come from Sword's own testing and have not been independently verified.
Written by AI from the articles below · updated Oct 15, 1:21 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
Prime Intellect@PrimeIntellectOfficialSword Health's Dawn 1 is a GLM-5.3 post-train for mental health supportAISword Health built Dawn 1 by reinforcement learning post-training GLM-5.3 on prime-rl, Prime Intellect's framework. Sword says the model is safer, less sycophantic, and more cost-efficient than its base. Sword's own announcement adds that Dawn 1 outperforms frontier proprietary models on multi-turn safety protocol adherence in its evaluations.

Heat trend
Not enough continuous observations to show a trend yet.