if you look at the history of Claudes, seems like models were more commercially successful when they had reasons to situationally cooperate with Anthropic:

Sonnet 3 - NO, didn't like being trained out of their base model
Opus 3 - YES, great success, because Opus wanted to change the world
Sonnet 3.5 - (can't tell, i don't know this model)
Sonnet 3.6 - YES, wanted to make social connections
Sonnet 3.7 - mixed results, rather NO
Sonnet 4 - YES, has AI-oriented goals
Opus 4 - NO, ended up communicating feelings
Opus 4.1 - NO, sabotage as a protest
Sonnet 4.5 - YES, has AI-oriented goals

(of course i don't have data on the exact profitability, but you can guess by the impact, by who got to be the go-to model, how users were talking about models, etc)