The price of Composer 1.5 has already surpassed that of Sonnet 4.6.
But does its capability truly exceed that of Sonnet 4.6?
Has anyone used or reviewed it in depth?
As long as there are no publicly disclosed clean benchmark of Composer 1.5 (except the one they disclosed when introducing the model which honestly does not tells much) it is really hard to tell. Each model has its own weak points while can shine on different fields. Personally I do not feel much difference, but really depends on the complexity of tasks and stack you are working with.
Regarding intelligence, I think the goal of Composer 1.5 is not only intelligence but also speed. The justification for its pricing is that it provides a balance of smart reasoning and fast performance, aiming to be on par with Sonnet 4.5 at the time Composer 1.5 was released.
Because of that, this model is a good choice for people who need both speed and Sonnet 4.5–level intelligence.
That’s my perspective on Composer 1.5.
3 Likes
No.
not it sucks. they are training it on our shared data… maybe try enabling your data sharing for it to get better ( also cheaper_