OpenAI shares plans for GPT-4.5, GPT-5 AI models: All you need to know

OpenAI's Sam Altman said that the company aims to unify its AI capabilities into a single, more advanced system with the GPT-5 model

OpenAI CEO Sam Altman has shared details on the company's plans for future artificial intelligence models. In a post on X (formerly Twitter), Altman announced that OpenAI will soon release GPT-4.5, internally referred to as Orion. He added that GPT-4.5 will be the company's last "non-chain-of-thought model," as OpenAI aims to unify its AI capabilities into a single, more advanced system with GPT-5.

"We want AI to 'just work' for you; we realise 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," he added.  READ: OpenAI plans to release new GPT-4.5 model in coming weeks, says Altman

Upcoming AI models from OpenAI: What to expect

GPT-4.5

 

Altman did not disclose detailed specifications of GPT-4.5, but he confirmed its internal codename as Orion. Reports from The Verge last year suggested that OpenAI was developing an AI model with this codename, with an OpenAI executive claiming it could be up to 100 times more powerful than the GPT-4 model. The report also stated that OpenAI has been training the Orion model using synthetic data provided by the o1 reasoning model. If true, GPT 4.5 could bring a significant boost in performance, despite being a mid-generation upgrade.

 

GPT-5

 

Altman revealed that OpenAI's next-generation model, GPT-5, will focus on unifying different AI capabilities by integrating the strengths of both GPT-series and o-series reasoning models. He said that the GPT-5 model will integrate a lot of the company's technology, including the o3 model's reasoning capabilities. He confirmed that the next-generation model will incorporate voice, canvas, search, deep research, and more features.  READ: What is Goose, Jack Dorsey's open-source AI?

 

He also shared how ChatGPT will evolve with GPT-5. Free-tier users will have unlimited access to GPT-5 at a "standard intelligence" setting, while paid subscribers will gain access to "higher intelligence" settings, with different levels of capabilities based on their subscription plan.

 

o3 reasoning models

 

OpenAI introduced the o3-series reasoning models in December and rolled out o3-mini for paid users in January. According to Altman, o3 will not be available as a standalone model but will be integrated into various AI-powered tools and agents. OpenAI has already deployed a "Deep Research" AI agent based on an o3-series reasoning model.