https://huggingface.co/baseten/GLM-5.2-Vision-FP8
This is a fascinating example of the fusion between "Moon-ViT(based on kimi-k2.6)" and "GlmMoeDsaForCausalLM(based on glm-5.2)" I’d like to request quantization for it.
It's queued, but the problem is it's too big and kind of a frankenstein, so not sure how it will work and when it will be quantized is unknown, because it needs to be handled manually
You can check for progress at http://hf.tst.eu/status.html or regularly check the model
summary page at https://hf.tst.eu/model#GLM-5.2-Vision-FP8-GGUF for quants to appear.
My understanding is that this model simply adds a vision encoder and a projector and updates the chat template, without any additional training of the language model itself. Therefore, if you swap the chat template (image add) of the existing GLM-5.2-gguf, you could reduce the conversion effort to just adding the GGUF files for the vision encoder and projector.