GLM CPU Promo Met With Sarcasm

DanGrover · x · 2026-07-13

This post quotes a promo for GLM 5.2 Colibri int4: marketed as a MoE model that runs entirely on the CPU without needing a GPU, targeting offline AI and low-barrier usage.

A commenter simply replied, "Narrator: it was not fast.", sarcastically implying that while being able to run is a selling point, the speed is likely far from ideal.

Related event: GLM CPU-only Inference Questioned as Speed Falls Below 0.2 tok/s(2 posts)→

Original post →

More from Fun

Fun channel →