Zhipu AI Taps Domestic Chips with Multimodal Model Launch
Zhipu AI has made significant strides in the field of artificial intelligence with the launch of GLM-5.3-Flash, its first natively multimodal model built for Chinese chips.
The model uses a hybrid architecture combining sparse and linear attention, which reduces compute requirements significantly.
This is particularly notable given the current export controls from the US that have restricted access to NVIDIA's most advanced chips for Chinese buyers.
GLM-5.3-Flash carries 320 billion total parameters but activates only 18 billion per token, making it a cost-effective option without sacrificing performance.
The model also boasts a long context window of 1 million tokens and native support for image and video inputs, making it genuinely multimodal from the architecture level up.