12GB coding guide

Best coding LLMs for 12GB VRAM

A 12GB GPU is a practical floor for local coding helpers. Prefer green Q4 fits for daily use; yellow models can be useful tests but are not the best default for long agent loops.

Open 12GB coding preset

Clean coding fits8
Stretch candidates8
Context profilecoding
Default quantQ4
Daily use

Pick green candidates for code review, small edits, and focused scripts.

Repo context

Longer repo context can push a green model toward yellow behavior.

Fallback

For large repo-scale coding, compare hosted/cloud models after local options.

12GB coding candidates

Clean fits first, then stretch tests.
16 candidates