- OpenAI CEO Sam Altman has laid out a roadmap for the release of new AI models: GPT-4.5 and GPT-5. He promised a simpler AI experience in future, with plans to unify OpenAI’s models.
Sam Altman, CEO of OpenAI, has unveiled the company’s next generation of AI models: GPT-4.5 and GPT-5.
In a post published to X on Wednesday, the tech boss said the company is working on simplifying its product offerings because OpenAI’s services have become too clunky.
“We realize how complicated our model and product offerings have gotten,” he said. “We hate the model picker as much as you do and want to return to magic unified intelligence.”
Altman said GPT-4.5, a model internally known as “Orion,” would be the company’s final non-chain-of-thought model, also known as models with “reasoning” capabilities.
He said GPT-5 would then unify the company’s series of reasoning models, like o1 and o3, with the GPT-series models into a cohesive system.
He said OpenAI’s “top goal” is to combine the models to create 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.”
The free tier of ChatGPT will also get unlimited access to GPT-5, while Plus and Pro subscribers will get access to versions of GPT-5 that run at higher intelligence levels. In a reply to the post, Altman said the release of GPT-5 and GPT-4.5 was “weeks/months” away.
This streamlined approach aims to remove complexity for users, replacing the current model picker with a more intuitive and versatile AI system that “just works,” according to Altman.
The timeline for GPT-5
OpenAI’s GPT-5 model has been hotly anticipated, but Altman has traditionally steered away from nailing down clear release dates for the prospective model.
When asked about plans to release the model in an interview with Lex Fridman in March last year, Altman said: “I think before we talk about a GPT-5-like model called that, or not called that, or a little bit worse or a little bit better than what you’d expect from a GPT-5.”
“I think we have a lot of other important things to release first,” he added.
The delays around GPT-5 have led some to question if the AI company is struggling to build its new frontier model.
Late last year, several reports suggested that OpenAI’s new AI model was behind schedule and costing more than expected. The reports kickstarted a debate in Silicon Valley about whether AI scaling laws had hit a wall.
Altman, however, has publicly dismissed claims that AI progress had stalled.
AI expert—and long-time OpenAI skeptic—Gary Marcus called Altman’s most recent post an attempt to put a “positive spin on what looks like a major retreat.”
“AI will soldier on, perhaps with new inventions, but pure scaling of data and compute didn’t even get us to GPT-5, let alone GPT-6,” he said in a post on X. “Even with the discovery of test-time compute, GPT-5 remains elusive.”
OpenAI’s new challenges
OpenAI has recently had its position as an AI leader challenged.
Chinese AI startup DeepSeek caused waves and tanked tech stocks last month with the release of its competing, but open-source, model R1.
The company’s chatbot shot to the top of app stores and shook investor confidence in the U.S.’s perceived dominance in the AI race.
Altman has already praised DeepSeek’s model as “impressive” and said OpenAI plans to “pull up some releases” in response.
He said that while he believed OpenAI would “obviously deliver much better models,” it was “invigorating to have a new competitor.”
The CEO has also said the model has prompted him to reassess his company’s stance on open vs closed-source AI.
“I personally think we have been on the wrong side of history here and need to figure out a different open-source strategy,” Altman said during an ‘Ask Me Anything’ session on Reddit. “Not everyone at OpenAI shares this view, and it’s also not our current highest priority.”
