Zhipu shipped GLM 5.2 to every coding tier with a 1M-token context and, notably, no benchmarks at launch.

On June 13, Zhipu rolled GLM 5.2 out to all four GLM Coding Plan tiers with a 1M-token window and two thinking modes, aimed squarely at long-horizon agentic coding. The standalone API, Z.ai access, and MIT-licensed open weights land next week. Deploying before publishing benchmarks is unusual, and reads as confidence, given GLM 5.1 already claimed roughly 94% of Claude Opus 4.6's coding performance. The MIT-licensed weights next week are the line item to watch after this week's Anthropic lesson.