OpenAI today introduced the next iteration of its AI models, GPT-4.5. GPT-4.5 is OpenAI's "best model yet" for chat purposes, with scaling in unsupervised learning that allows the model to better recognize patterns, draw connections, and generate creative insights.
GPT-4.5 has a more natural feel with an improved personality, and is able to better guide users through ideas and the steps that it takes to get to answers and ideas. It outperforms GPT-4o in almost every category, including everyday queries, professional queries, and creative intelligence.
Compared to GPT-4o, GPT-4.5 features a broader knowledge base, an improved ability to follow user intent, and a higher emotional quotient, which OpenAI says makes it useful for writing, programming, and solving practical problems. It could also hallucinate less.
OpenAI says that users can expect more warm and intuitive conversations more attuned to human collaboration because it is able to better understand what humans mean and it is better at interpreting subtle cues and implicit expectations.
Note that GPT-4.5 is not a reasoning model like o1 or o3-mini. It is better for general purpose use, and is built using pre-training and post-training. Reasoning models outperform GPT-4.5 at tasks like coding and math.
OpenAI is making GPT-4.5 available to Pro users starting today, with access coming to Team and Plus users next week. Education and Enterprise access will follow. Developers on all paid tiers will also get access to GPT-4.5 today.
GPT-4.5 supports file and image uploads and can work on writing and code, but it does not support Voice Mode, video, and screensharing.
