
Chinese lab Z.ai released GLM-5.2, an open-weight MIT-licensed model scoring 81.0 on Terminal-Bench 2.1, up from 62 on the prior version in weeks. It edges past GPT-5.5 on SWE-bench Pro. Experts foresee most advanced AI running on personal devices within years, threatening hyperscalers' massive datacenter investments.
Tap to vote and see what everyone thinks.
Summary by ByteBrief
Z.ai Launches GLM-5.2 With 1M-Token Context