The O1-PRO model of Chatgpt Pro is 140 times more expensive than the Chatgpt O3-mini. What is not clear is that it is much better

Paying $ 20 per month for accessing Chatgpt Plus did not seem too serious. The thing began to get interesting when Openai launched its chatgpt pro subscription at 200 dollars a month. But now there is an even more expensive version of these models, and this again shows that we are at the beginning of a worrying slope: that of AI prices.

What happened. OpenAI He has just announced That O1-Pro, the most powerful version of its reasoning model, is now available through its API. Until now it was only possible to enjoy it in a limited way through the subscription to Chatgpt Pro, but now the developers can also create services with it. The striking is not really that the company offers this version, but the price at which it is available.

140 times more expensive than O3-mini. As we said, O1-Pro is only limited to Chatgpt Pro users. If someone wants to use it intensively through the API, prices They are really high:

As can be seen, O1-Pro is 10 times more expensive than O1, and about 135 times more expensive than O3-mini, which is the version of the reasoning model that for example we can use (limitedly) in the free version of Chatgpt.

Gold Price Tokens. That cost imposes an important barrier for those who want to use this AI model for projects in which we introduce a lot of text and want to obtain a lot of text. For example, if we want to analyze a lot of documents and We saturate The input context window (200,000 tokens) and the output (100,000 tokens), we will have a cost of $ 90 … for a single question and an answer (an API request).

But is it really so good? Preliminary O1-Pro impressions in Chatgpt Pro They were not especially striking according to whom They have tried italso They have protested for one falling quality. According to Openai O1-PRO, it offers “responses that are better consistently”, and highlight that it is an O1 version that uses more calculation power to “think more and offer even better responses to the most complex problems.”

For asking not to be. IA companies are losing a money training their models, but they hope to recover it with a simple technique: increasingly faces. Data leaked in The Information also pointed out how Openai were already raising how an AI agent could replace a “human doctorate” to investigate, the cost would be $ 20,000 per month.

They seem faces, but maybe they don’t. Of course, the price of these subscriptions and these accesses to the API may seem expensive, but they can also be a true bargain. What matters here is the profitability that we can get out of these models of AI: if they help us generate many more income, we will end up having “robotic employees” who work 24 hours a day, seven days a week and also do so with a productivity and quality of theoretically extraordinary work. But of course, we will first have to use them like this … and ensure that these results are effectively valid, something that for now does not seem so clear.

In Xataka | AI agents are promising. But as in Tesla’s FSD, you better not take your hands from the steering wheel