OpenAI, the artificial intelligence (AI) company led by Sam Altman, is reportedly preparing to release GPT-5, the next generation of its multimodal large language model, in the coming months.
According to a Business Insider report citing two people familiar with the company, OpenAI could introduce GPT-5 sometime over the summer. The company has reportedly offered demos to enterprise customers, with one CEO who recently saw a version of GPT-5 describing it as “really good, like materially better.”
In March 2023, the company unveiled GPT-4. A refined iteration, known as GPT-4 Turbo, was introduced later that year.
“GPT-4 Turbo is more capable and has knowledge of world events up to April 2023,” OpenAI wrote in a blog post when it was introduced. “It has a 128k context window so it can fit the equivalent of more than 300 pages of text in a single prompt.”
In addition to developing GPT-5, OpenAI is also moving toward deploying its technology in humanoid robots through a collaboration with Figure AI. This partnership aims to accelerate Figure’s timeline by giving its humanoid robots the ability to process and “reason” from natural language.
The race to develop humanoid robots has intensified among big technology companies. Nvidia recently announced Project GR00T, a general-purpose foundation model for humanoid robots, along with a new computer called Jetson Thor and its Isaac robotics platform upgrades.
Earlier this year, OpenAI unveiled Sora, AI software that can create hyper-realistic one-minute videos based on text prompts. Sora is in the red teaming phase, where the company identifies flaws in the system. Sora leverages a neural network, which has been trained using video examples, to turn written scene descriptions into high-definition video clips that can last up to 60 seconds. Its superior performance compared to other AI video creation tools has captivated both the movie and technology sectors.
The release of GPT-5 may take a bit longer because the red teaming process hasn’t started, which is an important step. A red team is a group that pretends to be an enemy, attempts a physical or digital intrusion against an organization at the direction of that organization, and then reports back so that the organization can improve its defenses.