Run quantized versions. https://unsloth.ai/docs/models/glm-5.2