Guotai Junan emphasizes the release of the OpenAI o1 model, accelerating the AI computing power infrastructure.

Guotai Junan Securities released a research report stating that they are optimistic about the new models o1 and o1 mini from OpenAI in handling complex tasks such as science, mathematics, and programming at the inference end. They also believe that the continuous iteration of the new models will continue to drive the infrastructure development of AI computational power. They continue to recommend investment opportunities in AI hardware companies. Guotai Junan's main points are as follows: Investment recommendation: We are optimistic about the ability of OpenAI's new models o1 and o1 mini to handle complex tasks in science, mathematics, and programming at the inference end, and we also believe that the continuous iteration of the new models will continue to drive the infrastructure development of AI computational power. We continue to recommend investment opportunities in AI hardware companies. The strength of the o1-mini and o1-preview models lies in their ability to think, handle complex tasks, and better solve scientific, programming, and mathematical problems than previous models. They spend more time thinking before responding, just like humans. Through training, they learn to optimize the thinking process, try different strategies, and identify errors. The new models are more vertical in nature: the enhanced reasoning abilities of o1-mini and o1-preview may be particularly suitable for handling complex problems in fields such as science, programming, mathematics, and similar domains. For example, o1 can be used by medical researchers to annotate cell sequencing data, by physicists to generate complex mathematical formulas needed for quantum optics, and by developers in various fields to build and execute multi-step workflows. In categories that require more reasoning, such as data analysis, programming, and mathematics, o1-preview is more favored than GPT-4o. However, o1-preview is not favored in some natural language tasks, indicating that it is not suitable for all use cases. Compared to GPT-4o, the o1 model emphasizes reasoning, accuracy, and expertise, while o1 mini is a cost-effective and efficient reasoning model. The o1 series performs well in accurately generating and debugging complex code. Similar to how humans may take a long time to think before answering a difficult question, through reinforcement learning, o1 has learned to hone its thinking chain and improve the strategies it uses, simplifying complex steps into simpler ones. This process significantly enhances the model's reasoning abilities. Meanwhile, OpenAI o1-mini is a faster, cheaper, and more efficient reasoning model, particularly suitable for programming. As a smaller model, o1-mini is 80% cheaper than o1-preview, making it a powerful and cost-effective model for applications that require reasoning but do not require extensive worldly knowledge. Risk warning: AI-related capital investment falling short of expectations, AI-related policy implementation falling short of expectations, AI-related investment returns falling short of expectations, AI-related companies' revenues falling short of expectations.