March 20 news.OpenAI Yesterday (March 19) at X The platform released a tweet announcing that through developer API,Launched the upgraded version of o1 series "o1-pro".The company claims that it achieves "more consistent and better quality responses" through higher investment of computing resources.
The core benefit of o1-pro is a significant increase in computational resources to make it more comfortable with complex problems, but the new model is only available to select developers (who need to spend at least $5 in API services) and is priced well above comparable products.
According to the official pricing, "o1-pro" costs $150 per 1 million tokens (about 750,000 English words) to input, which is twice as much as GPT-4.5, and $600 per 1 million tokens to output (note: the current exchange rate is about RMB 4,339), a 10-fold increase compared to the normal o1 model. OpenAI emphasizes that the developer community's strong demand for stronger reasoning capabilities is the key driver for this upgrade.
A spokesperson for OpenAI said:
- o1-pro is an enhanced version of o1 that uses more computation to think about harder problems and provide better answers to the toughest ones. After receiving many requests from the developer community, we are excited to bring it to the API to provide more reliable responses.
TechCrunch reported that "o1-pro", despite its high-end positioning, has not performed particularly well in practice, and that since its launch on ChatGPT Pro in December 2024, users have found that it makes frequent errors when solving logic problems such as Sudoku, and is even "tripped up" by simple visual puzzles. "Tripping".
OpenAI benchmarks from late last year showed that the o1-pro was only "marginally ahead" of the regular o1 in programming and math, but with improved reliability.
Currently, the developer community's attitude towards o1-pro is polarized: supporters believe that its "stable output" is suitable for high-value scenarios, while critics question how well its high cost matches its real-world results. an OpenAI spokesperson said that the model will be continuously optimized to meet the needs of the profession.
