[evilduck]

8B models with larger contexts, or even 9-14B parameter models quantized.

Qwen2.5 Coder 14B at a 4 bit quantization could run but you will need to be diligent about what else you have in memory at the same time.