OpenAI CEO Sam Altman revealed on February 12 that ChatGPT’s next model upgrade will be GPT-4.5. Following that, the company will release the much-anticipated GPT-5 as a mixture of models including OpenAI o3—a slightly different form than expected—in a timeframe Altman described as “weeks/months” in a post on X (formerly Twitter). eWeek reached out to OpenAI to confirm release timetables for the new generative AI products but had not heard back as of publication.
“In both ChatGPT and our API,” Altman wrote, “we will release GPT-5 as a system that integrates a lot of our technology.”
GPT-4.5 will be the last ‘non-chain-of-thought’ model
Code-named Orion, GPT-4.5 has been rumored to be behind schedule and not as dramatic a generation leap as anticipated. Altman said it will be the last “non-chain-of-thought model,” as opposed to ‘reasoning’ models like OpenAI o1 and o3. The switch reflects a focus on the more advanced models and a desire to avoid confusion between the two lineages of names: the GPT varietals, and the ‘o’ varietals.
“We want AI to “just work” for you,” Altman wrote. “We realize how complicated our model and product offerings have gotten. We hate the model picker as much as you do and want to return to magic unified intelligence.” The model picker enables subscribers to choose which model ChatGPT calls.
All ChatGPT users will have access to GPT-5
GPT-5 will incorporate “a lot of our technology,” Altman wrote, possibly including multiple models beyond just OpenAI o3, which will not be released as a standalone product. OpenAI said o3 was in early access for safety and security testing in December. GPT-5 will be available to all tiers of ChatGPT users, including the free tier.
GPT-5 for ChatGPT voice, canvas, search, and deep research will be free. Those who subscribe to ChatGPT Plus can use GPT-5 “at a higher level of intelligence,” and Pro subscribers will also be able to access “GPT-5 at an even higher level of intelligence.”
Faith in OpenAI’s AI superiority was shaken earlier this year when DeepSeek released its ‘reasoning’ model, R1, which DeepSeek claimed earned some comparable or better benchmark scores.
See how ChatGPT fared against Google Gemini in our head-to-head comparison of the top chatbots.