Distilled Model's Vision Problem

<!-- SC_OFF --><div class="md"><p>Hi,</p> <p>I'm trying to use the vision part of Qwen3.6-35B but when I upload a text page from a webui (not llama own) and ask to OCR it, it

<!-- SC_OFF --><div class="md"><p>I removed mmproj file from models to remove vision and save my vram. But just curious, is this really don't affect its text ability? </p> <p>I

<!-- SC_OFF --><div class="md"><p>Since it came from a vision model i was wondering why it doesn’t ship with a mmproj file. Possible to get vision on 3.6 27b? </p> </div><!--

<!-- SC_OFF --><div class="md"><p>so recently i got to know that we can use vscode terimal + claude code + ollama models<br /> and i tried doing that it was great but im running

<!-- SC_OFF --><div class="md"><p>I am amazed by the development of locallm in the coding area. Right now im Testing Qwen3.6 27b and it works quite well, even tho this is not made

<!-- SC_OFF --><div class="md"><p>Given the recent usage reductions in claude, people have been looking for ways to save usage.</p> <p>Obviously a local model running on a single

<!-- SC_OFF --><div class="md"><p>I have a 2 DGX Spark cluster running dsv4. I have about 13gb free on the second one that I would like to allocate for a vision model to set up in

<!-- SC_OFF --><div class="md"><p>Trying to auto categorize a former employees photos from personal and work related. It’s a lot of photos and I don’t want the guy to loose