GLM-5.3-FlashX launches on B.AI with API and Web Chat support

The B.AI platform announced on Sept. 21 the launch of GLM-5.3-FlashX, a high-speed multimodal model developed by Z.AI. The model features a sparse architecture with 320B total parameters and 18B active parameters, achieving generation speeds of up to 200 tokens per second, which is approximately five times faster than the GLM-5.3-Flash version. It supports a 1M token context window and native video and file understanding. Developers can access the model via the B.AI platform or the web chat interface.