GLM 5.2 Hits 744B Parameters, Eval Approaches GPT-5.5

haider1 · x · 2026-07-03

The author predicts the emergence of a fable-tier open-source model within the next 5-6 months. They note that GLM 5.2 is already a 744B total parameter / 40B activated MoE model, with eval scores approaching GPT-5.5 and Opus 4.8. Given this trajectory, the prediction seems highly realistic.

Original post →

More from Models

Models channel →