I don’t understand how come a tool call that is <1000 tokens costing $0.05 per call when running sonnet 3.7 + MAX agent mode.
Also, previously sonnet 3.5 (in 0.45) was doing a lot more/better work, now, I’ve to use MAX with 3.7 to achieve the same things, but at a lot higher cost as compared with Cursor 0.45 + sonnet 3.6.
Question is why?
By quality degradation, I mean code quality in terms of ‘working code/fixes’.
Sonnet 3.5 on Cursor 0.45 was smashing bugs/problems, where all models in the latest versions of Cursor fail to solve simple things.
Not to comment that ~5-10% of my premium calls feels illogical/unneeded/wrong random guesses to try commands/fixes then go back and try something else (in case, I’d a situation where it kept rerunning the same command over and over again).
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.