OpenAI is still challenging the upper limit of LLM prices, and the API pricing of o1 pro is directly ten times that of the previous strongest inference model. I have been using the non API version of o1 pro for three months now, and it is really strong. I can run programming tasks almost once, as long as I can clearly describe the requirements and logic. Sometimes when I switch back to writing code on o1, the logic becomes disorganized. Unless I ask him to help me write SQL, I have already become accustomed to o1 pro. Although ChatGPT Pro is really expensive at $200 per month, I think it's worth it compared to the time and energy saved. I usually assign a task now, drink some water and go for a walk. After 3-5 minutes, it can give me high-quality results. Today, three months later, o1 pro is finally available in the API, and the price has also surpassed the sky: Input $150 per 1 million tokens and output $600 per 1 million tokens (but can save half through batch interface) By comparison, o1 is one tenth of it, and the o3-mini I commonly use in the API is even less than one percent of it. By comparison, I interact with him with dozens of questions every day, which saves me a lot of money. However, on the API side, I estimate that more people are using it for structured output capabilities or providing services to developers who have low demand but want to use it on demand. Additionally, developers on HackerNews have the following opinions: >Some people believe that although the price is high, the model performs well in tasks such as identifying code vulnerabilities, and its performance is sufficient to compete with high paying manpower. >Some opinions believe that these technical specifications are relatively mediocre and cannot reflect cutting-edge strength, while others believe that the performance of o1-pro has not significantly declined compared to previous versions and still has competitiveness. >Commentators cited OpenAI's terms of service, stating that the use of interfaces to copy or compete with its products is prohibited to prevent abuse. These comments are concerned that high pricing and strict usage restrictions may be a means to prevent the model from being used illegally or deviating from its intended purpose.