It's better on AIME '24, Multilingual MMLU, SWE-bench, Aider’s polyglot, MMMU, ComplexFuncBench while being much much cheaper and smaller.
and it's worse on just as many benchmarks by a significant amount. as a consumer I don't care about cheapness, I want the maximum accuracy and performance
As a consumer you care about speed tho, and GPT-4.5 is extremely slow, at this point just use a reasoning model if you want the best of the best.