No exl3 quants for Qwen3 VL 235B Thinking

[wypiki]

Qwen3 VL 235B would be a great alternative to GLM 4.6 on my 4xA6000 setup (more context, similar intelligence, multimodal) for me and my collegues. I just didn't find any quantized version for this for weeks now. Is it just a matter of time or is there a reason for this? Any experience qualitywise compared to GLM 4.6?

[Ph0rk0z]

I can tell you GLM is much better and I was disappointed in VL. it rambles and understanding of images is at the same level as pixtral/cohere. Try it in ik/llama.cpp if you want to see for yourself.