I legitimately dont know how to reply, bc by this point llms co-own all aspects of my life and jumps between gpt4->claude3->claude3.5->o1 have all been very noticeable
the problem is that last generation of the largest models failed to overcome smaller models on the benchmarks, see lack of new claude opus or gpt-5. The problem is probably in the benchmarks, but anyway.
regarding the prompt processing and token generation you are correct
it makes sense to benchmark them infependently since prompt processing is done in parralel for each token and is compute bound and token generation is sequential and bound by memory banwidth
Why is anything good objectively?
It is good for me because I can stay connected with my mom despite living in different country.
Also facebook group for my town is nice and useful, and my wife enjoys posting in instagram from our trips.
>... had very good universities with mathematics, physics, electronics, engineering. It had developed textile industry, built heavy carriers, tractors.