In the lead-up to Christmas, the field of artificial intelligence is once again making waves. Following Google's announcement of its advanced reasoning model o1, OpenAI quickly followed suit, revealing its next-generation model—o3—on December 20. This new model's launch has garnered widespread attention, showcasing OpenAI's significant advancements in reasoning capabilities and potentially altering the future landscape of AI development.

o3's Powerful Performance

OpenAI’s CEO Sam Altman stated during a live broadcast that o3 is "a very, very smart model." According to OpenAI's evaluation data, o3 performs exceptionally well across multiple testing areas. For instance, in software engineering assessments, o3 achieved an accuracy rate of 71.7%, nearly 47% higher than o1's 48.9%. In competitive mathematics testing, o3 recorded an accuracy of 96.7%, surpassing o1 by 15%. In tests involving biochemistry knowledge at the doctoral level, o3 also outperformed o1, with an accuracy increase of nearly 13%. These figures not only highlight o3's leading position in various domains but also indicate a breakthrough in AGI (Artificial General Intelligence) related assessments, with a top score of 87.5%, exceeding the human-level threshold of 85%.

Competition Between Google and OpenAI

Google's new model o1, upon its release, demonstrated its advantages in reasoning capabilities and transparency. o1 employs a slow-thinking reasoning approach that allows for deep visualization of the entire reasoning chain, particularly excelling in handling complex mathematical and programming problems. This new model has performed well in the Chatbot Arena large model evaluations, becoming a leader in the assessment rankings.

However, OpenAI's o3, once launched, attracted considerable attention. The testing results for o3 indicate that it surpasses o1 in multiple critical areas, demonstrating OpenAI's notable progress in AI reasoning capabilities. This back-and-forth competition between Google and OpenAI over reasoning models not only represents a contest of technical prowess but also reflects the strategic positioning of the two tech giants in the field of artificial intelligence.

Future Plans for o3

Despite the encouraging evaluation results for o3, OpenAI is not in a hurry to release it to the general public. Altman mentioned that the o3 series may not be available to ordinary users for some time, as there is a desire to establish a federal testing framework to guide monitoring and mitigate potential risks before the official release. He emphasized that ensuring the safety and reliability of the model is OpenAI's top priority, akin to the safety verifications required for new pharmaceuticals or aircraft.

According to OpenAI's plans, preview versions of o3 and o3-mini will first be made available to safety researchers, with a formal release expected in early next year. This strategy illustrates OpenAI's commitment to fostering technological innovation while actively seeking a balance between regulation and safety to address the challenges that AI technology may present.

How to Trade on uSMART

After logging into the uSMART HK APP, click on the search icon at the top right of the screen. Enter the stock code, such as "GOOG.US" to access detailed information, trading history, and trends. Click the “Trade” button at the bottom right, select the “Buy/Sell” function, and submit your order after filling in the transaction conditions.

（Source: uSMART HK）