Graham Neubig highlights how the newly released GLM-5.2 open-weights model delivers coding and agentic performance competitive with closed alternatives on text-only work, yet its massive scale means practitioners cannot simply spin it up locally and must hunt for efficient vLLM or SGLang setups instead.
OK, I tried GLM-5.2 and this is a good model.
Probably the first model good enough to eschew closed models from your workflow entirely (except if you need vision).
I know this won't run on your laptop, but what are the best current vllm/sglang serving recipes?
Even with strong benchmark scores on long-horizon tasks, the model’s size pushes users toward specialized inference frameworks where optimal configurations remain an active search rather than a settled recipe.
MIT-licensed release and API options at modest per-token rates position GLM-5.2 as a lower-cost path for non-vision workloads, though exact regional limits and long-term quota behavior stay unspecified in current reports.
Users are excited about GLM-5.2 as a strong open alternative to closed models and SGLang's practical cookbooks because they highlight performance gains and time-saving configs for serving LLMs.
No Digg Deeper questions have been answered for this story yet.
I just learned about these sglang cookbooks, which give the exact recommended settings for serving different models on different hardware https://docs.sglang.io/cookbook/autoregressive/intro
Incredible resource, thanks @ricklamers for sharing it!
open models are having **another** moment! GLM-5.2 is remarkably good
i've been using it with deepagents code (dcode), our coding agent built on our open harness, which makes it super easy to take new models for a spin
open model 🤝 open harness
@gneubig https://docs.sglang.io/cookbook/autoregressive/GLM/GLM-5.2 enjoy!
@ricklamers Incredible, thanks! I didn't know this existed 🤩
@gneubig @ricklamers these cookbooks saved me a ton of trial and error, love the practical configs
@sydneyrunkle facts — and the part people sleep on is how much the harness matters, a mid model on a solid harness will beat a great model on a janky one almost every time
