been messing with a tiny 0.6B model, trying to make it

Used Jev to pick the best of several attempts from a 0.6B model for tool calling, improving dev-set accuracy from 61% to 73% for under $0.50.

The original post

been messing with a tiny 0.6B model, trying to make it better at tool calling without paying for labels it answers each prompt a few times, have Jev pick the right attempt, train on those 61% → 73% on my dev set. spent less than $0.5 cents still early tho :) @typesafeai

Engagement when collected

Numbers are a snapshot taken from X when the case was added to the library (schema v1, collected 2026-09-19); they will not match today.