Writing code models that run on 24 GB

41 models with published weights that fit in 24 GB — a 24 GB card, or a Mac with 24. Counted at four-bit quantisation, weights only, leaving the machine about a third of its memory and a gigabyte for context: room for roughly 24 billion parameters. A mixture of experts is counted in full, because it is held in full even though only a few experts compute.

A 24 GB graphics card or a Mac configured with 24. This is the first size where the well-regarded mid-weight models — the twenty-something billion parameter class — run comfortably, and where a local model starts to be a real alternative to an API for daily work rather than a demonstration.

Models and agents for reading a codebase and changing it, rather than writing a snippet. The boards that matter here are the ones that run against real repositories — SWE-bench, SWE-rebench, Terminal-Bench — because a model can write a plausible function and still fail to make a test pass. Note whether what you are looking at is a model or an agent: an agent brings the harness, the file access and the loop, and is priced for it.

Wider

Writing code models Models that run on 24 GB