\n<\/p>\n
OpenAI has launched<\/a> a more powerful version of its o1<\/a> \u201creasoning\u201d AI model, o1-pro, in its developer API. <\/p>\n According to OpenAI, o1-pro uses more computing than o1 to provide \u201cconsistently better responses.\u201d Currently, it\u2019s only available to select developers \u2014 those who\u2019ve spent at least $5 on OpenAI API services \u2014 and it\u2019s pricey. Very pricey.<\/p>\n OpenAI is charging $150 per million tokens (~750,000 words) fed into the model and $600 per million tokens generated by the model. That\u2019s twice the price of OpenAI\u2019s GPT-4.5<\/a> for input and 10x the price of regular o1.<\/p>\n OpenAI is betting that o1-pro\u2019s improved performance will convince developers to pay those princely sums. <\/p>\n \u201cO1-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,\u201d an OpenAI spokesperson told TechCrunch. \u201cAfter getting many requests from our developer community, we\u2019re excited to bring it to the API to offer even more reliable responses.\u201d<\/p>\n Yet early impressions<\/a> of o1-pro, which has been available in OpenAI\u2019s AI-powered chatbot platform ChatGPT<\/a> for ChatGPT Pro<\/a> subscribers since December, weren\u2019t incredibly positive. The model struggled with Sudoku puzzles, users found, and was tripped up by simple optical illusion jokes.<\/p>\n Furthermore, certain OpenAI internal benchmarks<\/a> from late last year showed that o1-pro performed only slightly better than the standard o1 on coding and math problems. It did answer those problems more reliably, however, the benchmarks found.<\/p>\n<\/div>\n