GLM-4.6V GGUF (imatrix)
This is an imatrix-weighted GGUF conversion of zai-org/GLM-4.6V (a vision-capable GLM variant), packaged by mradermacher for llama.cpp-style local inference. The repository includes both an .imatrix.gguf file (useful if you want to generate your own quants) and a wide spread of downloadable IQ/Q quants.
The practical takeaway is that this is a large model. Even the smallest quant files are tens of GB, so you'll want to sanity-check disk, RAM, and runtime support before you download. If you're experimenting, it's usually best to start with a mid-range quant and a single test task (e.g. image captioning for a small set of images you control, or visual question answering) so you can validate that your mmproj setup and prompt formatting are correct before you invest in longer runs.
If you're comparing VLMs, keep your image set fixed and vary only one parameter at a time (quant level, sampling settings, or prompt format). That will tell you quickly whether quality changes are coming from the model or from runtime configuration.
Quick stats from the listing feed: pipeline: image-text-to-text · 1 like · 12199 downloads.
Source listing: https://huggingface.co/models?sort=modified