OpenAI unveils GPT-4.5 in research preview, offers early access to developers and Pro users

CHATGPT maker OpenAI on Friday (Feb 28) released a research preview of its latest model, GPT-4.5, granting early access to developers and subscribers of its US$200-a-month Pro plan.

The company did not specify when GPT-4.5 would be widely available, but said the preview would help it “better understand” the model’s strengths and limitations.

“We’re still exploring what it’s capable of and are eager to see how people use it in ways we might not have expected,” OpenAI said in a blog post.

The company described GPT-4.5 as having a “broader knowledge base, improved ability to follow user intent, and greater ‘EQ’ (emotional quotient)”, making it useful for tasks such as writing, programming, and problem-solving.

GPT-4.5 is expected to be an upgrade to 4o, OpenAI’s current mass-market model.

“We also expect it to hallucinate less,” it added, referring to artificial intelligence models generating false or misleading information.

The launch was widely anticipated. On Feb 13, OpenAI chief executive officer Sam Altman had hinted at the model’s roll-out in an X post, in which he shared the company’s road map but did not specify a release date.

“We will next ship GPT-4.5, the model we call Orion internally, as our last non-chain-of-thought model,” Altman wrote.

“Chain of thought” refers to a reasoning process where reinforcement learning algorithms allocate more time to tackle complex queries.

SEE ALSO

Singapore miles ahead in depth of technical knowledge: OpenAI research chief

OpenAI’s existing models that incorporate this technology include o1 and o3-mini.

Singapore has the highest proportion of ChatGPT’s paid subscribers globally on a per capita basis, with one in four people using the tool weekly.

On Feb 20, OpenAI announced that its weekly active users had risen 33 per cent to 400 million since December 2024.