Z.ai Ships GLM-5.2 With a Usable 1M-Token Context and No Launch Benchmarks

MarkTechPost·high signal

Z.ai released GLM-5.2 on June 13, a 744B-parameter MoE (40B active per token) with a 1,000,000-token input window — a ~5x jump from GLM-5.1's 200K — plus up to 131,072 output tokens and two effort modes (High, Max). It ships under MIT (weights pending the following week) and works day-one with eight coding tools including Claude Code, Cline, OpenClaw, and OpenCode via an Anthropic-compatible endpoint, but Z.ai published zero benchmark scores (no SWE-bench, Terminal-Bench, or Code Arena numbers) at launch.