The entire point of this post is that it's open weights, you can run it yourself and don't have to deal with the API issues. You really do have that choice.
You could subscribe to Anthropic/OpenAI for the rest of your life for the cost it would take to host GLM5.2 locally - you need 1.5TB of VRAM just for the weights
You could subscribe to Anthropic/OpenAI for the rest of your life for the cost it would take to host GLM5.2 locally - you need 1.5TB of VRAM just for the weights