The Ollama vision models (qwen3-vl, llama3.2-vision) cannot process images via the command line despite instructions in their documentation. When users provide image paths, the models treat them as text, not actual images. Testing confirmed no result even after restarting, updating, or reinstalling. The models claim they "can't view external files" or describe image content from paths.
#Mastodon #AI #Vietnam #TechProblem #Ollama #HệThốngAI #CôngNghệ #LỗiPhầnMềm



