I have this weird feeling that GPT-4 was the last True Model from OpenAI. It smells like 5 isn't going to be just a model but the infrastructure too. You go to ChatGPT and select "GPT-5" but instead get a 4.0 reasoning fine-tune with really good tools.

i'm presuming that GPT-5 won't have a new base model. when they release benchmarks; are they going to be based on raw outputs from a model? or is it going to be a whole "system" with web search and other weird tools and be "AgEnTiC" that they all wrap in a package called GPT-5?