4 ms·

Here's a recent comment [1] by an OpenAI engineer confirming that they do in fact make such trade offs between intelligence and efficiency. [1]: https://news.y

• int_19h 8mo ago
  That comment only says that they have a lot of different options for smaller & faster models that people can opt into. It doesn't say that they dynamically scale things up or down depending on demand.