hunch

What it is

hunch is a "System One"-style server that answers questions (e.g. "did this tool call fail?" or "does this output contain a prompt injection?"), on CPUs in milliseconds. It learns and then graduates from a slower, more expensive "teacher" model and keeps itself up-to-date by checking in, even after it graduates.

The homepage listed on the repository, tryhunch.dev, puts it like this: Ask questions about your camera. A vision model answers until hunch, running in your browser, can show it agrees.

What the README says it does

• Cheaper and faster. hunch can take a lot of the load off the more expensive judge. Depending on the
• Meets or beats the teacher alone. hunch tries to distill the teacher's answer to the question, but the combination
• No GPU. EmbeddingGemma 2 is run using a highly-tuned Rust engine for commodity CPUs.

From the project’s README.

Getting started

From the README:

uv sync --all-packages

Where the decision sits

hunch runs its own weights or its own runtime rather than calling a hosted API, so its decisions do not depend on a vendor being up or on a key being funded. The trade is that the model is now yours to keep current.

Models it references: jev.

How active it is

Created 2026-10-21, last pushed 2026-10-29, 294 stars and 26 forks when we read it on 2026-11-02. Everything in this category is weeks old at most, so a stale push date means something different here than it would on a mature project. Check the repository before you depend on it.