OpenAI CEO Sam Altman revealed last week that the company has decided not to release the o3 artificial intelligence (AI) model. In January, the company rolled out the o3-mini model to all ChatGPT users, and the more capable o3 model was expected to be launched this month. However, sharing details about its AI roadmap, Altman revealed that instead of a standalone o3 model release, the company will focus on a unified GPT-5 model that will integrate the capabilities of o3.
OpenAI to Launch GPT-4.5 Before GPT-5
In a post on X (formerly known as Twitter), the OpenAI CEO detailed the roadmap for GPT-4.5 and GPT-5. The company said that with the next two model releases, it will simplify its product offerings to eliminate the friction users experience while picking different models for different tasks.
Instead, Altman said that starting GPT-5, the company will focus on offering “unified intelligence,” a model that can access all the different OpenAI tools and can understand when to use reasoning and when not to. This way, the end user will not have to put much thought into which model is suitable for which task.
To achieve that, the OpenAI CEO said that the capabilities of o3 will be integrated into the GPT-5 AI model. This also means that the o3 model will not be released independently. The AI firm might be effectively ending the standalone release of the o-series AI models.
The company is also reinventing its tier-based offering strategy with GPT-5. Altman said that the GPT-5 AI model will be available to all ChatGPT users at launch, and those on the free tier will get unlimited chat access at the standard intelligence setting. This will be subject to the company’s abuse thresholds, likely to prevent any incidents of model distillations.
ChatGPT Pro subscribers will be able to run GPT-5 at a higher level of intelligence compared to the free tier, and the Pro subscribers will get the highest level of intelligence. The vague nomenclature around “intelligence” is likely because the company does not want to reveal the capabilities each version of the AI model will get. However, Altman did reveal that GPT-5 models will incorporate Advanced Voice, Canvas, Search, Deep Research, and more.
However, before any of this is introduced, OpenAI is planning the launch of the GPT-4.5 AI model, which will be the company’s last non-chain-of-thought (non-CoT) model. While Altman did not reveal any release date, he highlighted that the model could arrive in the next few weeks or months. Interestingly, he also revealed that GPT-4.5 is internally called Orion, a model that was highly rumoured to be the GPT-5.
