Only organizations with more than $10 billion in annual revenue need a security review under Z.ai’s GLM-5.3 license, and only when they offer the model as an external model service. Z.ai said GLM-5.3 is available for download, local deployment, fine-tuning and commercial use. The company said it spent two additional weeks on safety evaluations before publishing the weights because of the model’s advanced cybersecurity capabilities.

GLM-5.3 is the largest model in the GLM-5.3 line, with 744B total parameters, 40B active parameters, a 1M context window and 128K max output. The weights were live on Hugging Face, and vLLM, SGLang and OpenRouter said they had day-0 support or availability after the release.

Sign in to suggest edits

Key sources

  1. SOURCE@zai_org“Our most capable model for agentic coding and cyber defense”x.com
  2. SOURCE@zixuanli_“a security review is required only when an organization with more than $10 billion in annual revenue offers GLM-5.3 as an external model service.”x.com
  3. SUPPORT@vllm_project“744B total, 40B active, 1M context, 128K max output.”x.com
  4. SUPPORT@sgl_project“On real-world multi-turn agentic workloads, we measured 537.6 tok/s/user on NVFP4 and 413 tok/s/user on FP8”x.com
  5. SUPPORT@kimmonismus“FP8: 10–12× H100 or 8× H200”x.com
  6. SUPPORT@openrouter“Built for complex software engineering, long-horizon agents, and cybersecurity, with 1M context and configurable reasoning effort.”x.com
  7. SOURCE@ollama“Private. Fast. US and Europe hosted. No data retention.”x.com
  8. SUPPORT@baseten“The smartest open-weight model at 743B params”x.com
Markdown