OpenAI CEO Sam Altman has shared new details about the company’s upcoming AI models, GPT-4.5 and GPT-5. In a post on X, he acknowledged the growing complexity of OpenAI’s product lineup and emphasized the company’s goal of simplifying its offerings. “We hate model fitters as much as you do and want to get back to the magical unified [artificial] intelligence,” Altman wrote.
GPT-4.5: The Last Non-Reasoning Model
Altman revealed that GPT-4.5, internally known as Orion, will be OpenAI’s final non-reasoning model. The training process for Orion was completed in October last year, but it was later disclosed that OpenAI faced challenges due to high costs and limited data availability. The primary goal of GPT-4.5 is to unify the “o” models and GPT models, enabling systems to utilize all OpenAI tools more efficiently. This includes improving the model’s ability to decide when to engage in deeper reasoning and when to respond quickly, making it more adaptable to a wide range of problems.
GPT-5 and OpenAI’s Future Approach
OpenAI aims to release GPT-5 as a comprehensive system integrating various technologies, including the o3 model—the company’s most advanced AI to date. Altman confirmed that o3 would no longer be available as a standalone model. OpenAI introduced o3 in December and later launched o3-mini in January.
Once GPT-5 is launched, free ChatGPT users will have unlimited access to it in its default settings, though with some safeguards to prevent misuse, notes NIX Solutions. Meanwhile, Plus subscribers will experience GPT-5 with enhanced intelligence, and Pro subscribers will have access to an even higher level of capability.
Altman did not specify the exact release dates for GPT-4.5 and GPT-5 but mentioned they would arrive in the coming “weeks/months.” We’ll keep you updated as more details emerge.