I use Sonnet 80% of the time when I'm on claude.ai and on Claude Code my main agent is Opus and typically my sub-agents are Sonnet. I use this setup mainly because I genuinely don't know if anything I'm doing is complex enough to need Fable and I don't want to run up my usage too much. It's basically like trying to kill a gnat with a sledgehammer. Hell, at work Haiku almost always works just as well as Sonnet does an…
I start with Fable (not always but most of the time) and the reason why is because I find the smaller models don’t know enough to know when they don’t know enough. Like an LLM version of Dunning Kruger. Fable will often tell me when I’m wasting tokens on it. Opus wants to keep everything to itself because it doesn’t trust other models. Sonnet and Haiku never know any different. Actually Haiku will almost always say yes if I ask it if it wants to kick up to a higher model but it won’t offer that, but Fable will offer “this could be done by Sonnet so you might want to switch to that.” In the end I find I save a lot of money this way. Where Fable might get something done in $10 and like maybe 20-30 minutes including the back and forth with me, if I start the job instead with, say, Opus, I might be looking at $50 (on API) and more wall clock time. If I start it with Sonnet, costs are low enough that they won’t generally catch up to Opus costs but wall clock time will be a lot higher.