← Latest products

I finetuned 1.5B Qwen to near GPT-4o level bash generation perf

EasyCommand: open models, training data, and lessons from teaching small models to write shell commands. Purely a hobby side project to see how far I can push a really small model, using (mostly) automated training pipelines Original mention: https://news.ycombinator.com/item?id=49869735 There were a bunch of requests to release it.

Visit dirac.runexternal site

Reactions (1)

Similar products

by description similarity

  1. Werk

    simonask.github.io

    a simple build tool and command runner

    Hacker News93 pts

  2. Bash Dungeon

    github.com

    An educational dungeon crawler in the shell

    Hacker News164 pts

  3. FlashQwen

    github.com

    A from-scratch CUDA inference engine for Qwen3

    Hacker News5 pts

  4. Levi

    ttanv.github.io

    run AlphaEvolve on your Claude Code/Codex for dirt cheap

    Hacker News5 pts

  5. OpenGraviton

    opengraviton.github.io

    Run 500B+ parameter models on a consumer Mac Mini

    Hacker News13 pts