LaunchAI ModelsSeptember 8, 2026

Zai's GLM-5.3-Flash revealed as viral "Ox Alpha" model

GLM-5.3-Flash packs 320B parameters with 18B active per token and a hybrid linear/sparse attention architecture, leading real-world tasks on GDPval AA v2 at $0.09 per task. Zai ran the model openly, and unsloth has published a GGUF quantization with 74,544 downloads.

How this story unfolded

2 days · 0 reports · 3 community posts · from Sep 7

  1. Sep 7
  2. Sep 8
  3. Sep 9

More stories today

Open the live feed