OpenAI CEO Sam Altman announced via an X post Wednesday that the company’s o3 model is being effectively sidelined in favor of a “simplified” GPT-5 that will be released in the coming 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
“We want AI to ‘just work’ for you; we realize how complicated our model and product offerings have gotten,” Altman wrote. “We hate the model picker as much as you do and want to return to magic unified intelligence.”
To that end, OpenAI will no longer release o3 as a standalone model. Instead, GPT-4.5, which was code named Orion internally, will become the company’s final non-chain-of-thought model when it is released in the next couple weeks. GPT-5 apparently won’t be fully standalone either, but rather a model that “integrates a lot of [OpenAI’s] technology,” including o3 — a “rat king of different mediocre products,” as AI critic Ed Zitron puts it.
“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 wrote. “These models will incorporate voice, canvas, search, deep research, and more.”
He went on to explain that chat users will have unlimited access to the “standard intelligence setting,” so long as they don’t run afoul of the platform’s “abuse thresholds.” Altman did not elaborate on what those thresholds will be. As with all of OpenAI’s products, paying more will grant you greater levels of access. Plus subscribers will be able to use GPT-5 at a “higher level of intelligence” than free users while Pro subscribers will get access to an “even higher level of intelligence.”
This move comes as DeepSeek continues to eat OpenAI’s lunch following the release of its R1 reasoning model that offers performance equivalent to o3, but at a fraction of the cost and energy. Altman previously promised to “pull up some releases” to better compete with the Chinese startup and this news, in addition to the recent release of o3-mini as the baseline model for ChatGPT, appears to be part of that response.
