February 28th.OpenAI A blog post was published today (February 28) announcing the official launch of the GPT-4.5 AI model that claims to be its biggest and best chat model. As a research preview, GPT-4.5 will first be available to ChatGPT Pro users.
GPT-4.5 improves unsupervised learning by scaling up pre-training and post-training, making significant advances in understanding patterns, making connections, and generating ideas. Early tests show that interactions with GPT-4.5 are more natural, the knowledge base is broader, user intent is better understood, and there is a higher level of "emotional intelligence", which contributes to improved writing, programming, and problem solving, while reducing "hallucinations".
GPT-4.5 enhances unsupervised learning capabilities with broader knowledge and deeper understanding of the world by scaling up computation and data, as well as optimizing architecture and innovation.
In addition, GPT-4.5 has a wider knowledge base, higher "emotional intelligence", and better understanding of the user's intent; deeper knowledge and understanding, reducing the "illusion" phenomenon of GPT-4.5, and improving reliability.
GPT-4.5 employs new scalable techniques to train larger, more powerful models using data derived from smaller models, improving controllability, understanding of nuance, and natural dialog, in addition to excelling in writing and design, demonstrating greater creativity and aesthetic intuition.
GPT-4.5 is suitable for application scenarios that require higher emotional intelligence and creativity such as writing assistance, communication, learning, mentoring and brainstorming. It also demonstrates strong capabilities in intelligent body planning and execution, supporting multi-step coding workflows and automation of complex tasks, among others.
GPT-4.5 is a very large and computationally intensive model, more expensive than GPT-4o, and is not a replacement for it.OpenAI is evaluating whether to make GPT-4.5 available in the API in the long term.
An early leaked document reveals that "GPT-4.5 is not a cutting-edge model, but it is OpenAI's largest LLM (Large Language Model), and is more than 10 times more computationally efficient than GPT-4. It does not introduce 7 new cutting-edge features compared to previous inference versions, and underperforms o1, o3-mini, and deep research models in most preparatory evaluations."
1AI cites a blog post that describes how GPT-4.5 will be rolled out to Pro users first, with a rollout to Plus and Team users next week, and then to Enterprise and Edu users after that, though users can try it out on Microsoft's Azure AI Foundry platform for now.
