wandering why you haven't tried local recent cyber models from anthropic or openai instead of GLM unless they are not available locally?
Because your American models aren't open source, they only attack other people's websites, and there's no way to deploy them locally