https://Z.ai's GLM-5.3-Flash, a multimodal model with 320B parameters, has been released, outperforming GLM-5.2 at a lower cost. First tested as “Ox Alpha”, it uses a hybrid architecture and mHC for better scaling and cost efficiency. https://alternativeto.net/news/2026/8/glm-5-3-flash-tested-as-ox-alpha-released-with-hybrid-architecture-and-scaling-efficiency/