Comment by cjonas

you can easily run qwen locally in a pi harness (or similar) and just tell claude code to delegate all implementation work to it. claude will even setup the harness with the local model