Just two weeks after CEO Sam Altman shared the roadmap for OpenAI’s upcoming AI models, the company has officially launched GPT-4.5.
In a new system card, OpenAI says that GPT-4.5 is its “largest and most knowledgeable model yet.” Early testing of the model indicates that interacting with GPT-4.5 is more natural and that it is better suited for writing, programming, and solving practical problems than GPT-4o. Plus, the new model is less prone to hallucinations than its predecessors.
Among the biggest leaps forward for GPT-4.5 is the model’s ability to recognize tone and intent when interacting with humans. For instance, if you tell GPT-4o you’re going through a hard time after failing a test, it will provide you with a long list of ways you might try to improve your mood. Meanwhile, GPT-4.5 understands that you’re probably not looking for a numbered list but just want some sympathy, even if it’s from a chatbot.
GPT-4.5 is also much more succinct than GPT-4o. If you are less than thrilled to see dozens of paragraphs start generating each time you ask ChatGPT a question, you should look forward to less verbose responses when interacting with GPT-4.5.
The release of GPT-4.5 is obviously significant, but OpenAI notes that “GPT-4.5 is not a frontier model.” It has better world knowledge, improved writing abilities, and a more refined personality, but no “net-new frontier capabilities compared to previous reasoning releases.”
The company also says that despite its improvements in some areas, “its performance is below that of o1, o3-mini, and deep research on most preparedness evaluations.”
As of this Thursday, anyone with a ChatGPT Pro subscription can select GPT‑4.5 in the model picker on the web, mobile, and desktop. GPT‑4.5 will begin rolling out to Plus and Team users next week, then to Enterprise and Edu users the following week.
“With every new order of magnitude of compute comes novel capabilities,” OpenAI says at the end of its introduction. “GPT‑4.5 is a model at the frontier of what is possible in unsupervised learning. We continue to be surprised by the creativity of the community in uncovering new abilities and unexpected use cases. With GPT‑4.5, we invite you to explore the frontier of unsupervised learning and uncover novel capabilities with us.”
Following the launch of GPT-4.5, OpenAI has said that its next goal will be to unify o-series and GPT-series “by creating systems that can use all our tools, know when to think for a long time or not, and generally be useful for a very wide range of tasks.”
