Golden Finance reports that on the morning of the 18th, OpenAI launched the 9th day of technical sharing live broadcast, officially releasing the API for the o1 model, and significantly upgrading the real-time API to support WebRTC. Among them, the API for the o1 model has reduced the thinking cost by 60% compared to the previous preview version, and comes with advanced visual features; the audio cost for GPT-4o has decreased by 60%, while the price for the mini version has plummeted by 10 times. At the same time, OpenAI also released a brand new preference fine-tuning method, allowing large models to better grasp user preference styles through direct preference optimization algorithms.