OpenAI, the artificial intelligence research company, recently announced the launch of a new "reasoning" AI model called o1-pro within its developer API. OpenAI revealed that o1-pro utilizes significantly more computing resources than its existing o1 model, aiming to deliver "consistently better responses."
However, this high-performance model is currently only available to a select group of developers – specifically, those who have spent at least $5 on OpenAI API services. Even more striking is o1-pro's exorbitant pricing.
OpenAI charges $150 per million tokens for input text (approximately 750,000 words equals 1 million tokens), and $600 per million tokens for generated text. This price is double the input cost of OpenAI's another advanced model, GPT-4.5, and ten times the generation cost of the standard o1 model.
Despite the hefty price tag, OpenAI has high hopes for o1-pro's performance and believes its superior capabilities will justify the cost. An OpenAI spokesperson told TechCrunch: "o1-pro in the API is a version of o1 that uses more compute to think harder and provide better answers to the toughest questions. Following numerous requests from our developer community, we are excited to bring it to the API to provide more reliable responses."
It's worth noting that o1-pro was previously released in December to ChatGPT Pro subscribers on OpenAI's ChatGPT platform. However, early user feedback on o1-pro has been mixed. Some users reported that the model struggled with Sudoku puzzles and was even stumped by simple visual illusion jokes.
Furthermore, some internal OpenAI benchmarks conducted late last year showed that o1-pro only slightly outperformed the standard o1 model on coding and mathematical problems. However, these benchmarks also revealed that o1-pro was more reliable in answering these questions. OpenAI is attempting to improve AI model performance and reliability in complex tasks by investing in higher computing costs, although this is directly reflected in its high pricing strategy. Whether developers are willing to pay such a premium for this performance improvement remains to be seen.
Key takeaways:
- 🚀 OpenAI has released a more powerful AI model, o1-pro, designed for superior reasoning capabilities.
- 💰 o1-pro's pricing is extremely high, with input costs double that of GPT-4.5 and generation costs ten times that of the standard o1 model.
- 🤔 Early user feedback and internal testing show that o1-pro isn't perfect in all areas, but it is more reliable on coding and mathematical problems.
