GLM-5.2 builds for longer engineering tasks

Z.AI introduced GLM-5.2, emphasizing long-running coding work, a million-token context, and multiple reasoning-effort levels. Its technical publication describes architectural changes intended to make large contexts more practical, and the release uses an MIT license.

Context

A long input limit is useful only if the system maintains quality while using it. Teams should evaluate sustained work with intermediate mistakes, changing tool output, and older constraints that must still be respected. The final repository and its behavior provide stronger evidence than the amount of context accepted.