o3-pro正式推出！说声“嗨”就花了80美元，思考1+1用时16分钟

Core Viewpoint - OpenAI has launched its latest and most powerful reasoning model, o3-pro, while simultaneously reducing the price of the previous model, o3, by 80% without compromising its intelligence [1][12]. Group 1: Model Performance and Features - Official evaluations show that experts unanimously prefer the responses from o3-pro over o3 [2]. - o3-pro surpasses both o3 and o1-pro, becoming the most proficient OpenAI model in mathematics, science, and programming [5]. - OpenAI CEO Sam Altman expressed astonishment at the performance improvements of o3-pro compared to o3 [8]. - o3-pro is now available to ChatGPT Pro and Team users, with Enterprise and Edu users gaining access the following week [9]. - The API pricing for developers is set at $20 per million input tokens and $80 per million output tokens [10]. Group 2: Pricing and Cost Structure - The pricing for o3 has been significantly reduced to $2 per million input tokens and $8 per million output tokens, making it comparable to GPT-4o [13][14]. - The reduction in o3's price is attributed to a comprehensive optimization of the reasoning service architecture, allowing for more efficient inference [17][18]. - Despite the price drop, there are ongoing debates about whether o3 has truly maintained its intelligence level [19][21]. Group 3: User Experience and Performance Insights - Users have reported that o3-pro is capable of solving complex problems that other models struggle with, although it tends to have longer response times [30][32]. - A notable example includes o3-pro successfully completing a challenging word ladder task that other models failed to solve [37]. - In a comparison of tasks related to the human immune system, o3-pro demonstrated superior performance by identifying critical limitations that required comprehensive restructuring [41][43]. Group 4: Future Perspectives and Industry Insights - Altman shared insights on the future of AI, predicting significant advancements in cognitive agents and the potential for systems that can generate original insights by 2026 [54]. - The cost of AI is expected to approach that of electricity as data center production becomes increasingly automated [54]. - The industry is viewed as constructing a "brain" for the world, with its limits determined by innovative ideas [56].