PANews reported on December 18 that according to an OpenAI announcement, OpenAI announced the official launch of the o1 model API in the 9th day of the technology sharing live broadcast, and upgraded the real-time API, introduced WebRTC support, and reduced costs by 60%. The o1 model has advanced reasoning capabilities, supports function calls, structured output, and visual functions, and reduces thinking costs by 60% compared to the preview version. At the same time, the audio cost of GPT-4o has dropped by 60%, and the price of the mini version has been reduced to one-tenth of the original. In addition, OpenAI has introduced a new preference fine-tuning method, which uses a direct preference optimization algorithm to make the model more adaptable to the preferred styles of users and developers. Developers can also use the newly added Go and Java SDKs to more conveniently integrate OpenAI's functions.