Z.ai releases 744B-parameter GLM5.2 open-source coding model

A Chinese artificial intelligence company has released a new open-source model that demonstrates a rapid advance in performance for complex software engineering tasks, challenging the dominance of leading proprietary systems. Z.ai launched GLM5.2, a 744-billion-parameter model, on June 16, 2026.

The model is available under a permissive MIT license, allowing unrestricted commercial use, modification, and local deployment. A key feature is its one-million-token context window, a fivefold increase over its predecessor, GLM5.1, enabling it to process and maintain context across extremely long coding sessions and large codebases.

According to the company and independent evaluations, GLM5.2 performs competitively with top-tier proprietary models on several benchmarks critical to developers. The model scored 62.1 on SWE-bench Pro, a benchmark for real-world software engineering tasks, surpassing the score reported for OpenAI's GPT-5.5. On FrontierSWE, which measures long-horizon technical projects, GLM5.2 scored 74.4, trailing Anthropic's Claude Opus 4.8 by less than a single point.

The release marks a significant acceleration in the development pace of open-source AI. GLM5.2 is the...