Cailianshe, April 5 (Editor Niu Zhanlin) On Friday local time, OpenAI CEO Altman said that he will release o3 and o4-mini in the next few weeks and GPT-5 will be released in the next few months.

Altman posted on social media that day that the release of GPT-5 was later than expected, but there was a reason. The most exciting thing is that he has the ability to do a much better job of the GPT-5 than originally expected.

image

"We also found that it's much harder to integrate everything smoothly than we expected. And we want to make sure we have enough capabilities to support unprecedented demands we expected." This shows that OpenAI expects GPT-5 to be very popular after the release, and there may be a large number of users, so they need to prepare in advance.

Recently, with the new image generation function of ChatGPT, the number of active users, in-app subscription revenue, and downloads of OpenAI have all hit record highs. It is reported that the number of paid users of ChatGPT has exceeded 20 million, and at the end of last year this number was 15.5 million.

The technical specifications of the o3 and o4-mini are not yet public, but they are expected to fill the capability gap between the GPT-4 and the upcoming GPT-5. Industry observers believe that GPT-5 will have significant improvements in reasoning, planning and memory functions.

OpenAI previously mentioned when publishing details about GPT-5 that the company intends to provide users with unlimited chat access to GPT-5 with "standard intelligent settings", but will be restricted by "abuse threshold". ChatGPT Plus users will be able to run GPT-5 at a “higher smart level”, while ChatGPT Pro users will be able to run GPT-5 at a “higher smart level”.

Altman revealed earlier this year: "GPT-5 will integrate features such as voice interaction, canvas, in-depth search, in-depth research, and more." He mentioned a series of features that OpenAI has launched in ChatGPT in the past few months. "One of our important goals is to unify our big model by creating systems that can use all of our tools, be able to tell when deep thinking is needed, when not needed, and be widely used in various tasks," he added.

Meanwhile, OpenAI is facing growing pressure from competitors who have adopted an "open" approach to launching models, such as DeepSeek, an AI lab in China. Unlike OpenAI's strategy, these "open" competitors offer their models to the AI ​​community for experimentation and, in some cases, commercialization.

However, just recently, Altman also announced that OpenAI will release a "strong new open-weight model with inference capabilities" in the next few months. The model will have reasoning capabilities and will undergo additional security assessments.

"Special statement: The content of the above works (including videos, pictures or audio) is uploaded and published by users of the "Dafenghao" self-media platform under Phoenix.com. This platform only provides information storage space services.

Notice: The content above (including the videos, pictures and audios if any) is uploaded and posted by the user of Dafeng Hao, which is a social media platform and merely provide information storage space services."

[Editor in charge: Yu Lei PT032]