Zhipu has announced its intention to open source the 32B and 9B series of GLM (Generative Large Model) models, encompassing base, inference, and meditation models, all under the permissive MIT license agreement. Users can currently experience all models in this series via "z.ai". Additionally, the latest versions of the base and inference models have been synchronized on the Zhipu MaaS platform.
Notably, the inference speed of the GLM-Z1-Air/AirX-0414 model can achieve an impressive 200 Tokens per second. Furthermore, the cost of GLM-Z1-Air-0414 is remarkably lower, at only one-thirtieth of that of DeepSeek-R1.
