Exploring Next
Exploring Next — Ep 185 w/ Justy & Cody — z.ai's open source GLM-5 achieves record low hallucination rate and leverages new RL 'slime' technique
z.ai's GLM-5 achieves record-low hallucination rates using a novel 'slime' reinforcement learning technique, scaling to 744B parameters while undercutting competitors by 6x on pricing. The model features native document generation and Agent Mode capabilities for enterprise workflows.