OpenAI announced on Thursday the release of a research preview for its latest AI model, GPT-4.5, granting access to Pro users and developers worldwide. The company, backed by Microsoft, plans to expand availability in the coming weeks.
GPT-4.5 boasts improved pattern recognition, heightened emotional intelligence, and enhanced creative insights, though it does not feature advanced reasoning capabilities, according to OpenAI. The model also demonstrates a lower rate of “hallucinations”—incorrect or misleading outputs—compared to its predecessors, with a rate of 37.1%, down from 61.8% in GPT-4o and 44% in the o1 reasoning model.
CEO Sam Altman described GPT-4.5 as “a giant, expensive model,” citing GPU shortages as a reason for its phased rollout. OpenAI plans to introduce the model to ChatGPT Plus and Team users next week, followed by Enterprise and Edu users the week after.
Altman noted that the company is working to expand its computing capacity, with plans to add “tens of thousands of GPUs” in the coming days.
The new model supports file and image uploads and can assist with writing and coding tasks. However, it does not yet include advanced voice or video features.
