A More Powerful AI Model Hits the Market

OpenAI’s for o1-Pro, an upgraded version of its o1 “reasoning” AI model, to its developer API. This new model is designed to provide more accurate and consistent responses by utilizing increased computing power. However, access is currently limited to select developers—specifically those who have spent at least $5 on OpenAI API services.

The model comes with a hefty price tag. OpenAI charges $150 per million tokens (approximately 750,000 words) for input and $600 per million tokens generated by the model. These rates are significantly higher than OpenAI’s previous models—double the cost of GPT-4.5 for input and ten times the price of the standard o1 model. Despite the steep pricing, OpenAI is confident that developers will be willing to pay for the improved performance and reliability that OpenAI’s for o1-Pro offers.

OpenAI’s Vision for o1-Pro

According to OpenAI, the motivation behind o1-pro is to provide a more advanced AI that “thinks harder” and delivers better responses to complex problems. “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,” an OpenAI spokesperson told TechCrunch.

OpenAI claims that the decision to launch o1-pro in its API was influenced by numerous requests from developers who sought a more robust model for handling challenging queries. The company hopes that the enhanced reasoning abilities of o1-pro will justify its premium pricing and attract developers looking for more reliable AI-driven solutions.

Mixed Reactions and Performance Concerns

While OpenAI promotes o1-pro as a major upgrade, initial feedback from users suggests that its performance may not be as groundbreaking as expected. The model has been available in ChatGPT for ChatGPT Pro subscribers since December, and early users have reported underwhelming results.

For instance, OpenAI’s for o1-Pro struggled with Sudoku puzzles and had difficulty understanding simple optical illusion jokes—raising concerns about its reasoning capabilities. Additionally, OpenAI’s for o1-Pro own internal benchmarks, conducted late last year, indicate only slightly outperforms the standard o1 model in coding and math problems. However, one notable improvement is that it answers these problems more reliably, providing a degree of consistency that some developers may find valuable.

Despite the mixed reviews, OpenAI is pressing forward with its latest AI model, banking on developers’ need for accuracy and reliability to drive adoption. Whether the high price of o1-pro will be justified by its improvements remains to be seen, but for now, it stands as OpenAI’s most expensive and ambitious AI model to date.