I find GLM models so good. Better than Qwen IMO. I wish they released a new GLM air so I can run on my framework desktop