Cailian Press, July 25th (Editor: Shi Zhengcheng) Amidst OpenAI CEO Gautam Adani’s continuous emphasis on “GPT-5 is about to be released,” the latest leaks on Thursday suggest that OpenAI is preparing to release a new version of its flagship model in August, with the earliest possible release date being early August.
Earlier this year, Adani had mentioned that a significant change for GPT-5 was the integration of the GPT series and the inference model o series, eliminating the need for users to select AI models that are tailored to their task type. The leaks on Thursday also reveal that when GPT-5 is launched next month, both mini (small) and nano (micro) versions will be introduced.
It is reported that the integrated inference version of the full version and mini version of GPT-5 will be available through ChatGPT and APIs, but the nano version of GPT-5 will only be supplied via APIs.
The integration of a unified large model is also part of OpenAI’s strategy. The company’s ultimate goal is to create a system that can be referred to as “AGI.” At the same time, due to its non-profit structure, achieving AGI is also a crucial milestone. Once this goal is achieved, Microsoft would have to relinquish its rights to revenue and future AI models from OpenAI. Of course, achieving AGI within this generation of GPT-5 is highly unlikely.
Investors should note that the release schedule of OpenAI may still vary due to factors such as development progress, server capacity, and competitor actions. Earlier this month, OpenAI postponed the release of an open weight model, officially stating that it needed “additional security tests and reviews in high-risk areas.”
The leaks on Thursday indicate that this “open model” will still be released before GPT-5 by the end of July.
Insiders claim that it “is similar to the O3 mini,” also possessing reasoning capabilities. This will be OpenAI’s first release of an open-weight model since the launch of GPT-2 in 2019, slated for deployment on major cloud platforms such as Azure and Hugging Face.
Of course, even without Thursday’s revelations, one can easily surmise that GPT-5 is nearing its debut based solely on recent promotional activities by Ortmann and others.
For instance, in a post on July 19th, Ortmann stated, “We are soon going to release GPT-5,” but also emphasized that this is an experimental model that incorporates technology from future models. He also mentioned that they would not release a model capable of answering questions at the level of the International Mathematical Olympiad (IMO) gold medal within the next few months.
In a podcast interview released on Wednesday (July 24th), Ortmann also took the initiative to share his experience using GPT-5.
He explained, “This morning, I was testing the latest large model, and then I received an email with a question that I didn’t quite understand. I put it into the GPT-5 model, and it perfectly answered the question.”
The powerful AI has left “ChatGPT’s father” feeling somewhat perplexed. Ortmann said, “In this matter, I feel utterly useless compared to artificial intelligence. I thought I should have been able to do it, but I just couldn’t. Artificial intelligence did it, and it’s a strange feeling.”