Z.ai's GLM-5.2 Brings 1M-Token Context to Open-Weight Coding

Zhipu AI, the Beijing lab behind the international brand Z.ai, released GLM-5.2 on June 13, 2026, aimed squarely at coding and long-horizon agentic work. The model arrived first for subscribers on Z.ai's GLM Coding Plan, with MIT-licensed open weights and standalone API access scheduled to follow on June 16.

The headline specification is a 1-million-token context window, roughly five times the ceiling of GLM-5.1's 200,000-token limit, paired with support for up to 131,072 output tokens. According to a review by buildfastwithai, the model runs on a 744-billion-parameter Mixture-of-Experts architecture with around 40 billion active parameters per token, trained on 28.5 trillion tokens. Two selectable reasoning-effort levels, High and Max, let users trade throughput for analytical depth.

The important caveat is that Z.ai had not published any GLM-5.2 benchmark numbers as of launch day. The predecessor GLM-5.1 held a 1530 Elo on the Code Arena leaderboard, third globally at the time, and scored 58.4 on SWE-bench Pro according to the same review, but those figures do not carry over to the new release. Independent community evaluations will determine where GLM-5.2 actually lands once the weights are publicly available.

What may matter more than the benchmark question is the licensing. A 744-billion-parameter model released under MIT is not common. If the weights arrive as promised, any organization can fine-tune, self-host, or build commercial products on top of GLM-5.2 without royalties, which matters for teams that cannot route proprietary code through a hosted API. Interconnects.ai described GLM-5.2 as "the only open model mixing it up with OpenAI and Anthropic's latest models" in agentic coding work, though that is the author's framing rather than a settled benchmark conclusion.

Z.ai's Coding Plan Lite tier reportedly starts around $18 a month. For developers who need large-context reasoning over entire codebases and want an alternative to closed frontier pricing, GLM-5.2 is worth watching once the open weights land and third-party evaluations begin.