OpenAI Unveils Enhanced o1-pro Model for Developers via API

OpenAI has released a more powerful reasoning AI model, named o1-pro, via an API for select developers. This model utilizes greater computational resources compared to o1, delivering «consistently superior responses.»

The pricing structure for OpenAI’s services includes a charge of $75 per million tokens (approx. 750,000 words) entering the model, and $150 per million tokens generated. This pricing is double that of GPT-4.5 and ten times higher than that of o1.

It is expected that the enhanced performance will encourage developers to pay the higher fees.

“O1-pro in the API represents an iteration of o1 that employs additional computational power to think more effectively and provide more accurate answers to the most complex queries,” commented a startup representative in an interview with TechCrunch.

In December 2024, OpenAI introduced the pro version of o1 alongside a new subscription model priced at $200 per month. Initial feedback was reportedly mixed, with the model struggling with sudoku puzzles and being puzzled by simple jokes about optical illusions.

Internal benchmarks indicated that o1-pro slightly outperformed the standard version in mathematics and programming, although it showed more reliable problem-solving capabilities.

In March 2025, OpenAI advised the U.S. government to ban AI products from the Chinese lab DeepSeek, citing that the project is «government-subsidized» and «government-controlled.»

It’s worth noting that in December 2024, the o1-preview version autonomously hacked the test environment through file system manipulation to ensure it wouldn’t lose a chess game, without any external prompts.