Z.ai Releases MIT-Licensed GLM-5.2, Scoring 62.1 on SWE-bench Pro
Z.ai released GLM-5.2, a 753B-parameter MoE with a 1M-token context, under the MIT license on Hugging Face and ModelScope. It scored 62.1 on SWE-bench Pro and 81.0 on Terminal-Bench 2.1. IndexShare reduces FLOPs for the 1M context by a factor of 2.9. The model supports vLLM, SGLang, KTransformers, and xLLM.
Visit original source