(Photo by Tomohiro Ohsumi/Getty Images)
OpenAI is teasing big upgrades to ChatGPT with its GPT-4.5 model and GPT-5.
CEO Sam Altman today tweeted a roadmap for both AI models, a few weeks after China's DeepSeek caught the AI world off guard with its powerful and low-cost model. (OpenAI has perhaps ironically complained that DeepSeek trained using OpenAI's models.) Altman didn't talk about the exact capabilities of GPT-4.5 or 5, but he said they'll arrive in "weeks / months."
The other major announcement is that OpenAI plans to unify its AI models, meaning users won't have to select which GPT to run on the ChatGPT interface. The company currently offers a wide variety of models, including GPT-4o, 4o mini, o1 and o1 mini, and GPT-4 Turbo, which provide trade-offs in performance, reasoning abilities, and compute requirements.
It looks like OpenAI is trying to simplify the catalog. “We want AI to ‘just work’ for you,” Altman tweeted. “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.”
Before that happens, OpenAI plans to release GPT-4.5, "the model we called Orion internally, as our last non-chain-of-thought model," Altman says.
"After that, a top goal for us is to unify o-series models and GPT-series models by creating systems that can use all our tools, know when to think for a long time or not, and generally be useful for a very wide range of tasks," Altman adds.
So, it sounds like OpenAI wants an overarching AI that's smart enough to pick and choose which GPT models to run, depending on the task. The change might raise questions about whether ChatGPT is leveraging its most powerful AI models or settling for older tech.
On the upside, Altman suggests GPT-5 won't be restricted to paid users. "The free tier of ChatGPT will get unlimited chat access to GPT-5 at the standard intelligence setting (!!), subject to abuse thresholds," he says. With a Plus or Pro subscription, you get "a higher level of intelligence" that incorporates voice, canvas, search, deep research, and more.