Zhipu Launches GLM-5.3-FlashX Model With 200 Tokens/s Inference Speed

Zhipu has launched GLM-5.3-FlashX, a new AI model achieving inference speeds up to 200 tokens per second. The API and experience center are now open, with performance powered by 100,000 domestic chips alongside infrastructure and optimization investments. The release builds on the open-source GLM-5.3-Flash base model introduced on August 26, which features 320 billion total parameters, 18 billion active parameters, and a 1 million token context window. The model was previously tested anonymously under the codename "Ox Alpha."