My RULE of Thumb of choosing a models

This is mostly for setting up for expectation, since personally without LLM i could take 3 days (15 hours of active programming) to debug or implement a feature, but with Qwen 27B (even before Qwen 3.8) it take 4 hours.

And yes 0.5 tok/s is human, not accounting of deletion and pausing, that's also the reason i am fine leaving overnight code base wide analysis or fin tech and deep research.