Z.ai's GLM-5.2 promises large context, lacks benchmarks

Z.ai has launched its latest large language model, GLM-5.2, featuring a notable 1 million-token context window and offering two levels of thinking effort, High and Max. However, no benchmarks were released at launch.

The GLM-5.2's 1 million-token context window could greatly enhance coding capabilities by allowing it to handle entire code repositories in memory, moving beyond the constraints of previous smaller context windows. However, the lack of benchmarks makes it difficult to verify its performance claims.

Developers using Z.ai's products, especially those involved with large-scale projects or long-term coding tasks, may benefit from the new features but face uncertainty without benchmark data.

The absence of benchmark scores raises uncertainty about the model's performance and potential application issues. Without these metrics, stakeholders cannot accurately assess the improvements over previous models.