Zhipu AI Open-Sources GLM-5.3-Flash Multimodal Model
Zhipu AI has open-sourced GLM-5.3-Flash, a 320B-parameter native multimodal model (with 18B active parameters) previously tested anonymously under the alias "Ox-Alpha". The model matches the performance of Claude Opus 4.8 on key benchmarks while being offered at a fraction of the cost. This release demonstrates that highly capable, frontier-class multimodal models can be run cost-effectively on domestic Chinese hardware. It lowers the barrier to entry for developers requiring advanced coding, agentic workflows, and document processing capabilities. GLM-5.3-Flash features a hybrid architecture of sparse and linear attention, a 1-million-token context window, and native FP8 weights. It integrates visual capabilities directly into coding loops, allowing the model to observe user interfaces, render results, and complete complex office workflows.
## BACKGROUND
Multimodal models process multiple types of input, such as text, images, and code, simultaneously. Mixture of Experts (MoE) architectures, like the one used in GLM-5.3-Flash, keep computational costs low by activating only a fraction of the total parameters (active parameters) for any given task.