GPT-4.5 is released by OpenAI for professional users; broader access will follow.
Better pattern recognition, more original ideas, and less hallucinations are all features of the AI model.
For ChatGPT Pro users and developers globally, OpenAI has formally released a research peek of their most recent AI model, GPT-4.5. In the upcoming weeks, access will be extended. The release was verified Thursday by the Microsoft-backed AI company, which said the model has better emotional intelligence, more creative insight generating, and better pattern identification.
While ChatGPT Pro users will have immediate access to GPT-4.5, OpenAI intends to make ChatGPT Plus and Team subscriptions available the next week, with Enterprise and Edu users following the week after.
GPT-4.5 is “a giant, expensive model,” according to CEO Sam Altman, who also mentioned that the company’s capacity to launch the model concurrently across all user tiers was constrained by GPU constraints.
Next week, tens of thousands of GPUs will be added. It’s difficult to accurately forecast growth spikes that result in GPU shortages, but this isn’t how we want to do things,” Altman wrote in a post on X (previously Twitter).
GPT-4.5’s decreased hallucination rate—a problem for big language models that might produce inaccurate or deceptive information—is a notable improvement. According to OpenAI, GPT-4.5 has a 37.1% hallucination rate, which is higher than the 61.8% rate for GPT-4o and the 44% rate for the o1 reasoning model on X (previously Twitter).
