GLM 4.7 FlashX

GLM-4.7 FlashX extends Zhipu AI's Mixture-of-Experts foundation, building on a 30-billion parameter architecture where only 3 billion parameters activate per token during inference. This sparse activation design allows the model to deliver enhanced performance without the memory and compute demands of a dense 30B model. The architecture positions the model as a high-performance variant within the GLM-4.7 Flash family, optimized for scenarios that require both capability and efficiency.

The model targets practical deployment across diverse hardware environments, with benchmarked speeds ranging from 43 to 81 tokens per second on consumer-grade hardware like RTX 3090s and Apple Silicon machines. Beyond straightforward text generation, the model has been designed for creative writing, translation, long-context comprehension, and role-play applications where extended context windows provide meaningful advantage. The FlashX designation indicates it represents an enhanced tier within the GLM lineup, offering faster performance characteristics compared to the base Flash variant for users requiring higher throughput while maintaining the lightweight efficiency the series is known for.