OpenAI has introduced GPT-4.5, a research preview of its latest and most advanced AI model, which is now available to ChatGPT Pro users and developers worldwide. The model represents “a step forward in scaling up pre-training and post-training,” according to OpenAI.
GPT-4.5 advances OpenAI’s approach to AI development by focusing on unsupervised learning, a technique that enhances a model’s ability to recognize patterns, draw connections, and generate creative insights without explicit reasoning.
“GPT-4.5 is an example of scaling unsupervised learning by scaling up compute and data, along with architecture and optimization innovations,” OpenAI stated in its announcement. “The result is a model that has broader knowledge and a deeper understanding of the world, leading to reduced hallucinations and more reliability across a wide range of topics.”
The model was trained on Microsoft Azure AI supercomputers, and OpenAI states that it is expected to “hallucinate less” compared to its predecessors.
OpenAI notes that interacting with GPT-4.5 “feels more natural” and that the model has “improved ability to follow user intent, and greater ‘EQ.’” These advancements make it particularly useful for “tasks like improving writing, programming, and solving practical problems.”
The company also highlights GPT-4.5’s ability to integrate ideas naturally in conversations. “GPT-4.5 has a better understanding of what humans mean and interprets subtle cues or implicit expectations with greater nuance and ‘EQ,’” the company stated. The model also “shows stronger aesthetic intuition and creativity.”
According to OpenAI, GPT-4.5 has been evaluated against previous models, showing improved factual accuracy and reduced hallucination rates. In a comparative evaluation, OpenAI reported that GPT-4.5 scored a SimpleQA accuracy of 62.5%, higher than GPT-4o and other OpenAI models. Additionally, its hallucination rate was 37.1%, significantly lower than GPT-4o’s 61.8%.
OpenAI also shared data indicating that human testers preferred GPT-4.5 over GPT-4o in several areas, including “creative intelligence,” “professional queries,” and “everyday queries.”
As part of its release, OpenAI states that it conducted extensive safety evaluations on GPT-4.5. “Each increase in model capabilities is also an opportunity to make the models safer,” OpenAI said. The model was trained using “new techniques for supervision that are combined with traditional supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF).”
The company also referenced its Preparedness Framework, which it followed to “stress-test our improvements” before deployment.
GPT-4.5 is now available to ChatGPT Pro users on web, mobile, and desktop platforms. OpenAI announced that it “will begin rolling out to Plus and Team users next week, then to Enterprise and Edu users the following week.”
The model supports various capabilities, including “search, file and image uploads, and the ability to use canvas to work on writing and code.” However, OpenAI clarified that GPT-4.5 “does not currently support multimodal features like Voice Mode, video, and screensharing in ChatGPT.”
For developers, GPT-4.5 is also being made available through OpenAI’s API offerings, including the Chat Completions API, Assistants API, and Batch API. “Based on early testing, developers may find GPT-4.5 particularly useful for applications that benefit from its higher emotional intelligence and creativity—such as writing help, communication, learning, coaching, and brainstorming,” OpenAI stated.
However, due to its high computational requirements, OpenAI is evaluating “whether to continue serving it in the API long-term.” The company encourages user feedback to help guide this decision.
OpenAI acknowledges that GPT-4.5 is still in a research preview phase. “We’re still exploring what it’s capable of and are eager to see how people use it in ways we might not have expected,” the company stated.
“With every new order of magnitude of compute comes novel capabilities,” OpenAI added. “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.”
The post originally appeared on following source : Source link
