Back to directory

Newt

A Swift package that answers multiple-choice, graded and yes/no questions about a piece of text using an on-device language model on Apple Core AI, returning a probability for each possible answer, inspired by TypeSafe's Jev.

Newt

Newt is a Swift package that answers multiple-choice, graded and yes-or-no questions about a piece of text using an on-device language model through Apple's Core AI framework, returning a probability for each possible answer rather than free text. It requires an Apple silicon Mac running a recent macOS and Xcode, since Core AI does not exist on Linux, and works with an exported Qwen3 model that the developer prepares separately rather than one Newt bundles or redistributes itself.

The project's own documented status is candid about its limits: on ambiguous inputs Newt's returned numbers can swing sharply between prompt wordings, so the README recommends treating its output as a ranking rather than as a probability to threshold, and it publishes a full comparison of its answers against TypeSafe's Jev on the same published examples. It is added to a project as a Swift Package Manager dependency pinned to a branch.

More in Inference

See all 770
  • OpenDecision

    deepanwadhwa.github.io

    An open-source Python package that answers typed questions about application state or documents using a local natural-language-inference model, returning structured choice, score, yes/no or evidence-relation values instead of free text.

  • Lifeboat

    iterate.ai

    A local inference server that puts an OpenAI- and Anthropic-compatible API in front of open-weight language models running on hardware you control, from a laptop to a GPU cluster.

  • Claude S40

    github.com

    An unofficial Claude chat client for 2007-era Nokia Series 40 phones, built as a Java ME app plus a small Go relay server, with web search, reading mode and saved chats.