Will the cost of Gpt4o become cheaper after o1 models are released? Even with reasoning of o1 4o is still the best model for my use case, and it would be even better if the cost goes down?
o1 models are released, and I don’t see any cheaper pricing right now…
OpenAI has not reduced the price of any particular versioned model since 2022 (when they were open about those efficiency alterations to a named model); OpenAI have just produced downward-trending new version releases within the same series, informed by the model’s state of computation resource consumption at new model introduction.
Nothing on the near horizon says the same AI would be less expensive in terms of datacenter hardware or energy, or that OpenAI’s pricing strategy would diverge from the past. What can be predicted is that you won’t know until they announce availablity.
So, you can get your cheaper version utilizing less resources today – with gpt-4o-mini.
1 Like
What? Token prices have been consistently dropping since competition in the LLM market intensified, especially after Google entered the scene with substantial resources.
For example, a significant price drop for all GPT-4 and -3 models, both old and new, was announced at the last Dev Day event, though this may not have impacted the 32k version.
However, the part about the current prices being unaffected by the latest model release is correct.
That “significant price drop” was only for a new version of significantly-different capacity for understanding, and much faster token rate signifying the computation used on making them: gpt-4-turbo.
gpt-3.5-turbo-1106 also came in new dated versions on that date, where the input context pricing being lower also seems to correspond with input attention quality (and 0125, remaining steadfast in price, has been unpredictably better and worse)
GPT-4-0314 - same price as 1.5 years ago.
So I make the case that new dated version → corresponding price, where OpenAI hasn’t gone far outside of a linear relationship to the actual operation costs from efficiencies or dialed-back computation usage being implemented.
You get those gpt-4o coders writing you structured outputs and another level of trust hierarchy in the model, for no price increase, today… 
1 Like
Since this part of the conversation is apparently not about the cost of using GPT-4 dropping from $30.00 per 1M tokens (input) and $60.00 per 1M tokens (output) to $10.00 per 1M tokens (input) and $30.00 per 1M tokens (output) with the latest GPT-4-turbo version, I found this example:
But this is apparently not the point you’re trying to make.
I’m saying it’s expected that the prices for the 4o models will drop and likely soon.
Somebody who appreciates original GPT-4 said a while back
We’ll see - I like cheaper AI of the same quality, if there’s a place to vote.
Back in my day (GPT-2), they gave us the weights and let us run it on its own! Bah humbug! Get off my LLM lawn! 
I do see prices coming down in general, at least for the models I’m using. At the same time, there’s a race for compute to keep up with demand.
GPT-4 will likely drop in price when GPT-5 arrives…
1 Like
GPT-4 isn’t GPT-4-turbo isn’t GPT-4o isn’t GPT-4o-mini
GPT-4-0314 isn’t GPT-4-0613, GPT-4-turbo-1106 isn’t GPT-4-turbo 0125.
Price drops generally only come with new models. Likely quants, or wholly smaller models altogether.
Unless you count inflation I guess 
The original question is if the cost of using 4o will drop and the answer is: yes, most likely.
(…if you wait long enough for technology to progress)
Except: considering how you must answer a hypothetical question asked two years ago: “Will GPT-3 davinci become cheaper after GPT-4 is released?”
Try to answer that in hindsight, and you’ll see that generalizations can’t be made about whether particular models will even be around in the far-term, to be run on that more price-efficient hardware that may come about.
You’d have to say “the future will be more AI for less, but it will come with a different name” instead of addressing gpt-4o specifically.
I could impose, “where’s your evidence for these conspiracies over what OpenAI’s pricing plans are for its model” (in somebody else’s voice) – except I’m not an instigating bully.
@_j
You are saying that that prices for a model snapshot like gpt-4o-2024-08-06 will remain constant.
I am saying that that there will likely be a new snapshot with lower prices for the 4o model because that’s what we have been observing for the past year or two.
1 Like
You got it!
flowchart
100!
1 Like
They just cut the price of GPT 4o in half, from $5/million tokens to $2.50 like a month ago. Of course they did that because several of their competitors (Anthropic and Fireworks.ai) were at $3/million.
It’s difficult to predict what will happen pricing wise but I suspect what their competitors do pricing wise will play a role. They’re all in a race to the bottom.
1 Like