Run a 27B reasoning model from a 6 GB file on a laptop

Made by Prism ML

A heavily compressed build of a 27-billion-parameter model, squeezed to around 6 GB so it runs on an ordinary laptop or a single graphics card through llama.cpp.

The creator reports it keeps 98.2% of the full-size model’s benchmark average.

You could use it to…

Run a 27B reasoning model from a 6 GB download
Work through a maths or coding problem, fully offline
Try a compressed model that keeps 98% of the benchmark score

Can I use this?

Runs via llama.cpp

You'll need llama.cpp on a laptop or a single GPU · Setup needed

Worth knowing Every compression and quality figure here is the creator’s own, measured on its own set of benchmarks. Heavily compressed models can differ from the original in ways a benchmark average does not show. Reading images needs a separate extra file. Released under Apache 2.0. BuildTube has not run or verified this model.

Why it's here

Trending on Hugging Face

2,320 people have liked it on Hugging Face.
It was downloaded 3,766,691 times in the last 30 days.

Numbers from the snapshot taken 1 October 2026; not refreshed since.

Behind it

See the code on Hugging Face

BuildTube has not run or verified this project. Everything above is written from what the creator published.

Made this?