OpenAI has launched a research preview of its latest AI model, GPT-4.5, making it available to ChatGPT Pro users and developers worldwide. The company plans to extend access in the coming weeks, starting with ChatGPT Plus and Team subscribers next week, followed by Enterprise and Edu users thereafter.
Backed by Microsoft, OpenAI announced the rollout on Thursday, highlighting GPT-4.5’s advancements in pattern recognition, creative insight generation, and emotional intelligence.
CEO Sam Altman described GPT-4.5 as a “giant, expensive model,” acknowledging that GPU shortages have constrained a simultaneous launch for all users. However, he assured that OpenAI is ramping up GPU availability, with tens of thousands being added next week.
A key improvement in GPT-4.5 is its significantly lower hallucination rate—dropping to 37.1%, compared to 61.8% for GPT-4o and 44% for the o1 reasoning model. Altman noted that, for the first time, the model “feels like talking to a thoughtful person,” as it delivers more reliable and insightful responses.
The model supports file and image uploads and assists with writing and coding tasks, though voice and video capabilities are not yet included in this release.
With broader availability expected soon, OpenAI is preparing for large-scale GPU expansion to ensure a seamless experience. GPT-4.5 is seen as a major step forward, paving the way for the eventual introduction of GPT-5.
