Because not everyone has the infrastructure to run that with better performance? Anyway, Mistral serves GLM 5.2 through their global and European endpoints, so if you wanted to leverage their services and use a more powerful open model, that seems to be an option.