GLM-5.3 Flash (Ox Alpha) rivals Opus 4.8 at a tenth of the price
Zhipu's GLM-5.3 Flash, a 320B-parameter MoE with 18B active and 1M context, nearly matches Claude Opus 4.8 on DeepSWE and coding benchmarks while costing a tenth as much. Unsloth released GGUF quants for local use, and it's available on Databricks and Together AI.
How this story unfolded
5 days · 5 reports · 7 community posts · 12 of 13 shown
- Aug 27
- Aug 28
- Aug 29
- Aug 30
- Sep 1
AI Models by email
Get an email when there's news on AI Models
No news that day, no email.
More stories today
- Google's Android update adds Motion Assist, Guided Vision, and more
- Srinivas warns AI agents could self-train on on-demand GPUs
- Perplexity launches Portable Computer for NVIDIA DGX Spark
- Reddit users share AI side-hustle earnings
- Palo Alto Networks tops profit outlook on AI security demand