[Zhipu Launches Native Multimodal Large Model GLM-5.3-Flash, Running 'Near Opus' Capability on Domestic Chips at Only One-Tenth the Cost] According to the official website of Z.ai, Zhipu has launched the native multimodal large model GLM-5.3-Flash, featuring a total of 320 billion parameters and 18 billion active parameters. It significantly outperforms GLM-5.2 in encoding and intelligent benchmarks and approaches Claude Opus 4.8, with costs reduced to approximately one-tenth. The model utilizes a hybrid architecture of sparse and linear attention, Manifold-Constrained Hyper-Connections, and 30 trillion multimodal datasets, reducing long-context reasoning costs to about one-third of GLM-5.3. On the Artificial Analysis Intelligence Index v4.1.1, it achieved a performance score of 57 at approximately $0.045 per task.