OpenAI unveiled the o1-pro model — powerful but very expensive

OpenAI has unveiled the powerful o1-pro version of its AI model in an API for developers. The company claims that the model uses more computation than the regular o1 to provide consistently better answers. It is currently available only to selected developers who have spent at least USD 5 on OpenAI API services, which is very expensive.

TechCrunch writes about it.

Why the model is so expensive to use

The company charges USD 150 per million tokens (approximately 750,000 words) entered into the model and USD 600 per million tokens generated by the model. These prices are twice as expensive as GPT-4.5 from OpenAI and 10 times more expensive than the usual o1.

OpenAI is convinced that the improved performance of o1-pro will encourage developers to pay such money.

"O1-pro in the API is a version of o1 that uses more computing to think harder and provide even better answers to the hardest problems," the OpenAI representative said.

He added that the company received many requests from the developer community, so the model was introduced in the API.

Nevertheless, the first impressions of o1-pro, which has been available on the OpenAI ChatGPT AI-based chatbot platform for ChatGPT Pro subscribers since December, were not positive. The model couldn't cope with Sudoku puzzles found by users and stumbled over simple jokes with optical illusions.

Internal tests conducted last year showed that the new model is only slightly better than the standard o1. It is slightly better at coding and maths tasks, although it does so more reliably.

As a reminder, OpenAI has updated the ChatGPT app for iOS. Now, iPhone and iPad users can use the chatbot as the search engine by activating the special extension in the Safari browser.

We also wrote that OpenAI has acknowledged, through internal tests, that the GPT-4.5 model makes up facts almost every time you ask a question. The Grand Parlance Model is "hallucinating", confidently inventing information in 37% of cases.