GLM-5.2

Z.ai announced GLM-5.2 on June 16, 2026, positioning it as a flagship model built for long-horizon tasks. The release centers on a solid 1M-token context window that the team says is engineered to remain stable under real coding-agent workloads, not merely large. GLM-5.2 ships under an MIT open-source license with no r Architecturally, Z.ai introduces IndexShare, which reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9× at 1M context length, and improves the MTP layer for speculative decoding to increase acceptance length by up to 20%. The model supports multiple thinking-effort levels

The Z.ai developer documentation release notes corroborate the June 16, 2026 GLM-5.2 entry, confirming three core capability claims from the official blog. Specifically, the docs list 1M lossless context with reduced context drift and goal forgetting in complex tasks, open-source SOTA performance on coding and long-hor The same release-notes page also documents subsequent versions, GLM-5.3 on August 18, 2026 and GLM-5.3-Flash on August 26, 2026, but per scope these distinct variants are not transferred to the GLM-5.2 record. The 2026-06-16 GLM-5.2 row is the authoritative, version-exact match for this subject and serves as a first-pa