Just two days ago (Feb 11, 2026), Zhipu AI (aka Z.ai) dropped GLM-5 - a massive 744B-parameter Mixture-of-Experts (MoE) model under full MIT license. Trained entirely on Huawei Ascend chips (no NVIDIA in sight), it's already topping open-weight leade...

Markdown supported · ⌘↵ to post
No responses yet.