1 day ago
3
OpenAI this week announced an improved version of its o3 generative AI reasoning model, saying it performed better on benchmarks and is designed to spend more time thinking on difficult problems. The new model, called o3-pro, is now available for Pro and Team users in ChatGPT and API users, and will come to Enterprise and Edu users next week.
In its release notes, OpenAI said the model is similar to o1-pro, which users favored for its math, science and coding skills. It incorporates improvements from the newer o3 model, which can search the web and use more reasoning skills.
"Because o3-pro has access to tools, responses typically take longer than o1-pro to complete," the company said. "We recommend using it for challenging questions where reliability matters more than speed, and waiting a few minutes is worth the tradeoff."
Kevin Weil, OpenAI's chief product officer, posted on X that the company is also dropping the price of o3 in the API by 80%, and that rate limits for o3 will be doubled for Plus users.
(Disclosure: Ziff Davis, CNET's parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
OpenAI continues to build on its latest rounds of updates to its large language models, which rolled out in April. CEO Sam Altman has teased the larger release of an "open-weights" model, a kind of middle ground between open- and closed-source models that provides more insight into how the models work. He doubled down on that teasing this week, posting on X that it would be exptected later in the summer, but not in June.
"[O]ur research team did something unexpected and quite amazing and we think it will be very very worth the wait, but needs a bit longer," he said.
