OpenAI has officially launched o1-pro, the most advanced and expensive iteration of its o1 “reasoning” AI model, now available for developers through their API. This state-of-the-art model is designed to offer enhanced performance, harnessing significantly greater computational power than its predecessor, o1, to deliver “consistently better responses” to complex queries.

As it stands, access to o1-pro is restricted to a select group of developers who have invested at least $5 in OpenAI’s API services, making the model exclusive and, as implied, quite costly. The pricing structure is staggering—OpenAI is charging developers $150 for every million tokens (approximately 750,000 words) input into the model and an astonishing $600 for every million tokens generated. This represents a price tag that is double that of the current GPT-4.5 for input and a staggering tenfold increase over the regular o1 model.

In making this bold move, OpenAI is betting on o1-pro’s superior capabilities to attract developers willing to commit significant resources. According to an OpenAI spokesperson, “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. After receiving many requests from our developer community, we’re excited to introduce it to the API for more reliable responses.”

However, the reception of o1-pro has not been entirely favorable. Although the model has been available through OpenAI’s chatbot platform, ChatGPT Pro, since December, initial feedback has raised some eyebrows. Users reported that o1-pro struggled with even token puzzles such as Sudoku and was perplexed by basic optical illusion jokes, raising concerns about its reliability for practical applications.

As AI continues to evolve, the question remains: will the steep investment in o1-pro be justified by its capability to solve complex problems better than its predecessors? Only time and further user feedback will tell if OpenAI's gamble pays off in the competitive world of artificial intelligence.