you think companies are not focusing on efficiency but in reality they are both almost fully correlated. if you make a model more efficient, you get to make it

you think companies are not focusing on efficiency but in reality they are both almost fully correlated. if you make a model more efficient, you get to make it do more reasoning and we know that the upper limit for reasoning is very very high. we have real proof - GPT 5.6 Luna is 50% cheaper than 5.5. And this keeps repeating.