AnyDecisionModel brings answers with probabilities to Apple devices, from the model's own scores
A Hugging Face post on 29 October introduced two open Swift packages for Apple platforms. AnyDecisionModel answers yes-or-no, multiple-choice and rating-scale questions with typed results, deriving each answer from the model's probabilities rather than from generated text: a yes-or-no question returns a probability, and a choice returns the selected option with a probability for every option. Backends include MLX, Clef decision heads, Core AI, llama.cpp, Jev and Ollama.
The post is explicit that its confidence value, based on the entropy of the probabilities, measures how concentrated they are, not whether the answer is correct. The companion package, AnySpeechRecognizer, runs NVIDIA Parakeet and Alibaba Qwen3-ASR speech recognition locally through Core ML. Both need Swift 6.2 and macOS 26 or iOS 26 or later, and so far have been tested on macOS 27 only.
A probability you can read is a start, not a guarantee. Before you set a threshold on it, check on your own data how often answers at each probability are actually right.
Drafted with AI from the original source, then checked and edited by Sridhar Mukkandi. Spotted a mistake? Write to hello@tarkika.com and we'll correct it here.