Zhipu Open Source Native Multimodal Model GLM-5.3-Flash

Zhipu officially launched GLM-5.3-Flash, the first native multimodal model in the GLM-5 series. Its AA comprehensive intelligence index of 57 points is on par with Claude Opus's 4.8, but its price is only 1/40th of the latter. The model adopts a sparse + linear attention hybrid architecture, nearly halving the number of activation parameters and layers, and significantly reducing the cost of long contexts. It natively integrates visual encoding capabilities, supporting collaboration between code, browser, and GUI.