Which model do you actually use for daily work? Not architecture, not refactors, just regular stuff
With Sonnet 5.0 out, I'm noticing it feels more about cost than actual intelligence gains. And honestly, 4.6 has been struggling with stuff that used to be easy for it.
I'm not asking about complex architecture decisions, big refactors, or hard algorithmic problems. I mean the regular day to day tasks. Quick fixes, small features, routine debugging, the stuff that makes up most of your actual workday.
What do you use for that? Opus on low effort, Opus on medium, 4.7 on low, something like that?
And I'm not talking about API costs, I mean usage limits, since I'm guessing most people here are on a plan, not pay per token. What's actually the best cost benefit for daily driving, considering the limits you get?