That sounds like the GPU labor theory of value that was debunked a century ago.

No, its the fraud theory of charging for usage that is unaccountable that has been repeatedly proven true when unaccountable bases for charges have been deployed.

The one-shot models aren't going away for anyone who wants to program the chain-of-thought themselves

Yeah, if they are charging for some specific resource like tokens then it better be accurate. But ultimately utility-like pricing is a mistake IMO. I think they should try to align their pricing with the customer value they're creating.

Not sure why you didn’t bother to check their pricing page (1) before dismissing my point. They are charging significantly more for both input (3x) and output (4x) tokens when using o1.

Per 1M in/out tokens:

GPT4o - 5$/15$

O1-preview - 15$/60$

(1) https://openai.com/api/pricing

My point is that "cost fairness" is not a thing. Either o1 is worth it to you or it isn't.

If there's a high premium, then one might want to wait for a year or two for the premium to vanish.

It’s really unclear to me what you understood by “cost fairness”.

I’m saying if you charge me per brick laid, but you can’t show me how many bricks were laid, nor can I calculate how many should have been laid - how do I trust your invoice?

Note: The reason I say all this is because OpenAI is simultaneously flailing for funding, while being inherently unprofitable as it continues to boil the ocean searching for strawberries.

Eh it’s not worth it to me because it’s unfair.