OpenAI (United States) - OpenAI has just announced the launch of a research preview of its latest and most powerful artificial intelligence model, GPT-4.5. This model, available to Pro users and developers worldwide, represents a significant advance in scaling up pre-training and post-training for AI.

GPT-4.5 marks a milestone in the evolution of artificial intelligence, improving the ability to recognize patterns, establish connections, and generate creative ideas without the need for complex reasoning. Through unsupervised learning, GPT-4.5 significantly enhances its ability to handle large volumes of data, offering a broader knowledge base and a better understanding of the world.

Early tests have shown that interacting with GPT-4.5 feels more natural and intuitive. The model now has an improved ability to follow user intent and a more refined "emotional intelligence," making it ideal for tasks such as improving writing, programming, and solving practical problems. One of the most notable improvements is the reduction of hallucinations, a common issue in previous models.

This version is being shared as a research preview so that OpenAI can assess the model's strengths and limitations. The company is still exploring its capabilities and looks forward to seeing how users will leverage GPT-4.5 in ways that have not yet been anticipated.

The model was trained on Microsoft Azure AI supercomputers, using a significant increase in computing power and data volume. As a result, GPT-4.5 exhibits greater accuracy in world models and more refined intuition, providing more reliable responses and better performance across a wide range of topics.

This new model is one of OpenAI's greatest innovations to date, representing an important step forward for the future of AI, with more power, reliability, and efficiency.