For Z.ai's GLM-5.3, post-training is all you need

Z.ai released GLM-5.3 on Friday, a post-training upgrade to its GLM-5.2 model, which the company says is much better at complex coding and long-horizon tasks. The release follows Hugging Face's use of GLM-5.2 after OpenAI's attack on its infrastructure, when restrictions on cybersecurity AI prevented the use of Anthropic's Claude models.

Z.ai /tag/z-ai/ Z.ai's GLM-5.2 model was the tool Hugging Face turned to in the wake of OpenAI's attack on its infrastructure https://www.thestack.technology/openai-we-hacked-hugging-face/ when it was unable to use Anthropic's newest Claude models, thanks to government-imposed restrictions on cybersecurity AI. On Friday Z.ai released GLM-5.3, which it described as basically a souped-up version of 5.2. "Scaling post-training is all we did for GLM-5.3," the company said in a blog post, explaining that the heavy lifting that catapulted GLM-5.2 into the enterprise spotlight https://www.thestack.technology/opus-killer-glm-5-2-demand/ allowed the company to simply well, it's not simple https://pytorch.org/blog/a-primer-on-llm-post-training/?ref=thestack.technology fine tune the base model with additional data to create the new one. "Compared with GLM-5.2, it is much better at complex coding and long-horizon tasks," the company said. Get the full story: Subscribe for free Join peers managing over $100 billion in annual IT spend and subscribe to unlock full access to The Stack’s analysis and events. Subscribe now https://www.thestack.technology/membership/ Already a member? Sign in https://www.thestack.technology/signin/