Before the bigger models, the better it was. But now they didn't use Parameter scaling, in GPT-5, they unified a bunch of other models like reasonings, router, fast etc... Basically, instead doing what you say, it decided on it own what's the best. I guess OpenAI isn't winning the AI arm race, and getting closer to ASI anytime soon... unless they decided to change thier me to CloseAI and make it open source( Yeah..it is better than Claude(it uses 70$/million tokens) since it uses 10$/million tokens. Also, it kinda doesn't suck at making apps that much now...)