It is fully localized

The MWS Cloud team, part of MTS Web Services, announced the deployment of the GLM-5.3 model in its own cloud infrastructure. According to the company, this is the first placement of the model in a Russian cloud. Access to it is provided through MWS GPT Model Hub, with user requests and data processed within Russia. The cost of use will remain at the level of the previous GLM-5.2 version.

GLM-5.3 was developed by Chinese Z.ai. The model is distributed with open weights and is primarily intended for programming and performing complex multi-stage tasks. Application scenarios include writing and verifying code, finding errors, working with repositories, automating processes, and creating AI agents.

GLM-5.3 is based on the GLM-5.2 architecture, and the developer attributes the quality improvement to post-training scaling. Open weights were published two weeks after the release – Z.ai used this time for additional testing and security checks.

Placing the model in MWS Cloud allows companies to use it without transferring processed information outside of Russia. Computations are performed directly within the Russian cloud infrastructure.

According to developer estimates and independent tests, GLM-5.3 shows results comparable to a number of modern closed models. Its performance in programming and long-term agent scenarios is particularly notable.