ChatGPT / Depositphotos
OpenAI CEO Sam Altman announced plans to release new AI models: thus, GPT-4.5 will be available in «the coming weeks», while GPT-5 is scheduled for «the following months».
OPENAI ROADMAP UPDATE FOR GPT-4.5 and GPT-5:
We want to do a better job of sharing our intended roadmap, and a much better job simplifying our product offerings.
We want AI to “just work” for you; we realize how complicated our model and product offerings have gotten.
We hate…
— Sam Altman (@sama) February 12, 2025
GPT-4.5 is the model that was hiding under the name Orion (previously thought to be o1/o3) and it will be the last one to not have a «chain of reasoning». With the release of GPT-5, all OpenAI models will become «one», and artificial intelligence with the ability to reason will simply become another common feature of the company’s products.
«The most important goal for us is to unify the o-series and GPT-series models by creating systems that can use all of our tools, know when to think long and when not, and generally be useful for a wide range of tasks,» Altman says.
In fact, OpenAI will no longer offer o3 as a separate model (only within GPT-5). It should be reminded that the company first presented it in December, and launched the o3-mini version in January.
After the release of GPT-5, Altman promised to make it available in a free version of ChatGPT — with users getting «unlimited access to chat» with «standard intelligence settings». Plus and Pro subscription holders will instead be able to use the model at «higher levels».
Source: Engadget, The Verge