[nelsnelson]

Completely ignores externalized costs, and focuses entirely on purely end-user retail costs to operate, not even vendor internal operational costs. Can't even find the words "energy" or "electricity" or "scale" in the post. Whatever point this person is making, it is of such dramatic limitation that I am going to contentedly ignore it. Point people at this all you like, Juho Snellman. I for one will merely ignore you.

Also laughably excludes this one from openai's pricing details:

o1-pro-2025-03-19 Price per 1M tokens Batch API price -- Input: $150.00, Output: $600.00

And this doesn't even address quality. Results quality is also explicitly ignored. I personally find most results from cheaper models to be far, far worse than any results I find using search prior to the LLM content flood. But of course, that's 1) subjective, and 2) completely impossible to conduct any analytical comparison now since indexed search has been so completely ruined by SEO and LLM junk. Yet another externalized cost for which accounting is completely impossible, but is likely to have immeasurably negative impacts on the world's ability to share information.