GLM-5.2 Fast

GLM-5.2 Fast runs Z.ai's GLM-5.2 on a throughput-optimized path that delivers two to three times standard generation speed at the same quality bar. The underlying model, released in June 2026 under an MIT license, is a 744B-parameter mixture-of-experts design that debuted as the top-ranked open-weight model on the Artificial Analysis Intelligence Index. Agentic coding and terminal work are where it shines, with strong showings on SWE-bench Pro and Terminal-Bench, and its context window stretches past one million tokens with output up to 131K. The Fast tier keeps those strengths while cutting time to completion, a good match for interactive coding agents, large-codebase analysis, and reasoning-heavy pipelines where latency matters.