GitHub radar
Free Local Model For Lead And Comment Sorting
A lot of leads and comments? And requests via an API cost a fortune? This model works on the text and multiple questions and gives one single answer with probabilities: is it a hot lead? Is it spam? Is it toxic? Local, and for free, the small version is really fast!
Processes text once and provides several static outputs (questions) about it (e.g., “Hot Lead/Visitor?”, “Spam?”, “Who should deal with this?”). Outputs a numerical value rather than text (“Spam”, “0.89”). Runs locally on your Mac. Free. Small model runs extremely quickly.
Why a vibe-coder should care
While request-based lead/comment classification can get expensive when done at scale, this model allows you to do the same thing locally and for free, and the small model runs incredibly quickly. However, it’s less accurate than the first model for unseen data (0.80 vs 0.86 on their test set). Low stakes? Perfect. High stakes? Not so much.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up kev-4b on my Mac and build a lead triage on top of it: hot or not, show me a sample answer
kev-4b needs a Mac with an M-series chip and 32 GB of RAM. The 0.5B version works with less memory, but accuracy is lower.
Open on GitHub▌ More finds