Sam Altman, chief executive officer of OpenAI, during an interview at Bloomberg House on the opening day of the World Economic Forum (WEF) in Davos, Switzerland, on Tuesday, Jan. 16, 2024.
Chris Ratcliffe | Bloomberg | Getty Images
OpenAI on Thursday announced the research preview of GPT-4.5, a general-purpose large language model. It will initially be open to software developers and people with ChatGPT Pro subscriptions.
The model presents inaccurate information less frequently than predecessors, OpenAI said in a blog post.
“Early testing shows that interacting with GPT‑4.5 feels more natural. Its broader knowledge base, improved ability to follow user intent, and greater ‘EQ’ make it useful for tasks like improving writing, programming and solving practical problems,” OpenAI wrote. GPT-4.5 demonstrated a lower hallucination rate than OpenAI’s GPT-4o and o1 models in one test, the company said in a report accompanying Thursday’s release.
At the same time, the new model won’t be the state of the art in benchmarks, OpenAI co-founder and CEO Sam Altman said in a Thursday X post, describing the model as having “a magic to it i haven’t felt before.”
The release nevertheless represents progress for one of the most highly prized startups in the world. Last month CNBC reported that OpenAI was in discussions on a $40 billion funding round at a $340 billion valuation.
With over 400 million weekly active users, OpenAI intends to bring GPT-4.5 to those with ChatGPT Plus and Team subscriptions next week, the company’s research lead, Alex Paino, a member of the company’s technical staff, said on a livestream.
ChatGPT Edu and Enterprise subscribers will gain access the following week, Paino said.
GPT-4.5 is a “giant, expensive model,” Altman wrote in Thursday’s X post. “we really wanted to launch it to plus and pro at the same time, but we’ve been growing a lot and are out of GPUs. we will add tens of thousands of GPUs next week and roll it out to the plus tier then.”
GPT-4.5 is available in preview through Microsoft’s Azure AI Foundry, CEO Satya Nadella announced on X. Microsoft has invested over $13 billion in OpenAI and has incorporated the startup’s models into its products. Microsoft also helps to provide computing resources to OpenAI.
The release of the artificial intelligence model comes a month after markets reckoned with the efficient approach employed by Chinese lab DeepSeek. The market capitalization of Nvidia, whose chips are widely used among AI model developers, declined nearly $600 billion in a single day.
Two weeks later, Altman wrote in an X post that the company wants to get better at sharing a roadmap publicly.
Altman said GPT 4-5 would be coming out, followed by a GPT-5 that would draw on a wider variety of OpenAI’s technology. The company also offers so-called reasoning models that make many calculations at the time when people submit queries. GPT-4.5, by comparison, would be the company’s “last non-chain-of-thought model.”
WATCH: Distillation and the future of the Softbank, OpenAI partnership
