Openai has moved token: the same day as presented O3 pro The price of O3 cut 80 %. It may only be segmentation, but very possibly it is also an advance of the price war that starts when the rivals already touch their performance and internal efficiency allows us to serve Tokens for cents.
The new step is like this (prices per million Tokens):
ENTRANCE |
EXIT |
|
---|---|---|
OpenAi O3 |
$ 2 (before $ 10) |
$ 8 (before $ 40) |
OPENAI O3 PRO |
$ 20 |
$ 80 |
Gemini 2.5 Pro |
$ 1.25 |
$ 10 |
Claude 3.7 Sonnet |
$ 3 |
$ 15 |
Deepseek R.1-0528 |
$ 0.55 |
$ 2.19 |
That price narrowing comes because quality already converges in practice: almost all models “reason” comparable in common tasks, and hardware and inference efficiency improvements (on the side of OpenAi and the rest) reduce the marginal cost of serving each Token.
The result is manual: The price becomes the main lever and segmentation appears as a containment dike —How the creation of O3 Pro, designed for customers willing to pay a lot due to low latencies, huge contexts or formal audits.
For developers this is fertile land. 2 dollars a million tokens, AI ceases to be a luxury expense and looks like a CDN: Payment for use, without permanence contract.
Yes tomorrow Gemini either Deepseek They offer better value for money, change is reasonably simple in infrastructure and Prompts. The blocking effect is transferred to the layer that surrounds the model (tools, data, user experience), not the model itself.
At the contrary end of the cost spectrum, Apple has integrated its foundational models directly into operating systems. Any app can be invoked at home, free and without counters from Tokens No invoices at the end of the month: AI becomes as invisible as GPS or camera. It is another function.
The play Remember the emergence of WordPress in the early 2000s. At that time, publishing online required or a lot of knowledge or a lot of money to pay someone to have it. WordPress loaded that barrier and democratized the publication. Twenty years later we gave the xataka legend to its creator precisely because of that.
And now what? Possibly, an adjustment downward the margins in the base models and a migration of the business towards higher level services: observability, management of the management Fine Tuningprivate data accommodation, vertical agents … exactly what happened with the cloud when the price of the gigabyte sank and the value moved to the management layers.
The “magic” of AI is blurred, but its usefulness – now much more affordable – opens the door to a much wider market. Just what Openai needs.
In Xataka | O4-mini is much more than another model of AI. It is the Tesla Model 3 of OpenAI
GIPHY App Key not set. Please check settings