Z.ai releases GLM-5.2
Z.ai released GLM-5.2 for long-horizon agent tasks, with a stable one-million-token context window and MIT-licensed open weights.
What happened
Event details
GLM-5.2 focuses on sustained coding and agent workflows rather than context length alone. Z.ai says the model improves long-horizon execution, adds multiple reasoning-effort levels, and introduces IndexShare to reduce sparse-attention indexing cost at one-million-token context. The model is available through Z.ai services and as downloadable weights for local deployment.
Assessment
Why it matters
Combining a practical one-million-token context with open weights gives developers another frontier-scale option for long-running coding and agent workloads without regional licensing restrictions.
Availability
Access notes
GLM-5.2 is available on Z.ai and through the GLM Coding Plan; its model weights are published for local deployment under the MIT license.