Image processing in chat doesn't work (Qwen3.6-27b

[gthvidsten]

I have installed Qwen3.6-27b in ollama (batiai/qwen3.6-27b:q6). Qwen3.6 is supposed to support images, but when I try to attach an image to a chat message I get the following error:

Error: {"error":{"code":400,"message":"Multimodal data provided, but model does not support multimodal requests.","type":"invalid_request_error"}}

What am I doing wrong here?

[PurpleDoubleD]

this one is not on you, the error comes from the model and not from lu. the qwen3.6 build you pulled is text only, it has no vision part, so ollama refuses the image. for images use a build that can see, like gemma 4, llava, or a qwen vision model. 2.5.5 also now shows a plain message saying the model cannot read images, instead of the raw json.

[RedlineGT]

when running through built-in engine and a vision capable model how does the built-in engine wire up the mmproj file?

[PurpleDoubleD]

the lu engine picks up a vision file when it sits next to the model and is named like the model with .mmproj.gguf at the end, which is how the app's own downloads save it. one you drop in by hand under its original name isn't found, rename it to match the model and it is.