Should we use a non-thinking model for code after using a thinking one for plan? (Agentic coding)
I usually use Qwen3.6 27B (slow as heck on my RX 6800 but it works) for plan and Qwen3.6 35B A3B for the coding. But I was thinking the other day if I should remove the thinking from the code model. Is there a way to disable the thinking from the cod…