Ollama's decision capability allows local models to classify text, evaluate images, answer yes-or-no questions, and score content against custom rubrics without requiring an external internet service.
Built into Ollama version 0.35.0 and later, this runtime feature evaluates inputs directly on your machine through models such as Nimble, Tev1, Clef, and Clef Flash. Clef models add image evaluation starting in version 0.35.1. The system outputs probabilities for choices, boolean states, or rubric levels. Ollama notes that high confidence ratings reflect strong answer preference rather than a guarantee of correctness, and local execution requires hardware capable of running the chosen model weights.
How you can use it
Install a supported Ollama version and download a decision model such as Nimble. You can pipe a text file into its question command and supply choices or rating levels; local requests need no service key. Use Clef with version 0.35.1 or later for images, and check that your computer can hold the model.
