OpenAI has now created so many new models that even AI experts are finding it hard to keep track, but the company seems to have plans to rectify the situation.
OpenAI has said that it’ll unify its two model naming series into one with the launch of GPT-5. OpenAI currently has the reasoning ‘o’ series of models, and also the GPT-x series of models with multimodal abilities. OpenAI says that GPT-5 — its latest and most capable model — will combine the abilities of both these model series names into a single series.
“We know that our model names have become quite complex to follow along, with o3, o4-mini, GPT-4, GPT 4.1 and so on,” Romain Huet, OpenAI’s Head of Developer Experience recently said. “We’re truly excited to not just make a net new great frontier model, we’re also going to unify our two series. So the breakthrough of reasoning in the O-series and the breakthroughs in multi-modality in the GPT-series will be unified, and that will be GPT-5. And I really hope I’ll come back soon to tell you more about it,” he added.
OpenAI’s first models had been called GPT-1, GPT-2 and GPT-3. It was with GPT-3.5 that OpenAI became mainstream — GPT-3.5 powered ChatGPT back in late 2022. In March 2023, ChatGPT had released GPT-4, which had appeared to be a big jump in capabilities over GPT-3.5. OpenAI had then released GPT-4-Turbo, which was a faster and cheaper version of GPT-4.
In 2024, OpenAI began appending an o to its GPT models, releasing GPT-4o, GPT-4o Mini and GPT-4o Pro. In September 2024, OpenAI had released o1, which was its first reasoning model. In December, OpenAI had skipped o2, and announced o3, which was again a big step up in capabilities with its reasoning. Since then, OpenAI has released o3-mini, o3 and o4-mini and o3-pro. OpenAI had then reverted to its old naming scheme, and released GPT 4.5 in February this year.
OpenAI now says that GPT-5 will be the next iteration both for its o series of models, and its GPT-4.5 series, and will combine both reasoning and multi-modal abilities. GPT-5 is widely anticipated, because GPT-4.5 hadn’t felt as much of a jump over existing models, and the AI community is now hoping that GPT-5 will be an extremely capable model. And it now appears that it’ll also end up simplifying the naming schemes for OpenAI’s models, and be its clear flagship model that users can use when they want the best possible OpenAI model to answer their questions.