Qwen3-VL

Ollama announced on October 14, 2025 that Qwen3-VL, described as the most powerful vision language model in the Qwen series, is now available on Ollama's cloud. Ollama said the models will be made available locally soon.

Ollama listed the model's capabilities as visual agent operation of PC and mobile GUIs, including recognizing elements, understanding functions, invoking tools and completing tasks; visual coding boost for generating Draw.io, HTML, CSS and JS from images and videos; advanced spatial perception for judging object positions, viewpoints and occlusions, with stronger 2D grounding and 3D grounding for spatial reasoning and embodied AI; and long context and video understanding with native 256K context expandable to 1M, handling books and hours-long video with full recall and second-level indexing.

Further listed capabilities include enhanced multimodal reasoning in STEM and Math with causal analysis and logical, evidence-based answers; upgraded visual recognition from broader, higher-quality pre-training; expanded OCR supporting 32 languages, up from 19, described as robust in low light, blur and tilt, better with rare or ancient characters and jargon, and improved at long-document structure parsing; and text understanding on par with pure LLMs through text-vision fusion.

The model is run with the command ollama run qwen3-vl:235b-cloud. Ollama said users can prompt the model with a message and image paths, use multiple images, and drag and drop images to auto-type the file path. It listed examples including flower identification with a question about cat toxicity, menu understanding and translation, and basic linear algebra.

Ollama said its cloud can be used for free to get started with the full model through its CLI, API, and JavaScript and Python libraries, and that the model can be accessed directly through ollama.com's API. It said its OpenAI compatible API endpoints support the chat completions, completions and embeddings endpoints.