OpenAI launches GPT-4.5 with API access for developers

OpenAI has introduced GPT-4.5, a research preview of its most advanced language model to date. This release marks a significant step in the development of AI through scaled unsupervised learning, which enhances the model's ability to recognize patterns, generate creative insights, and reduce hallucinations. GPT-4.5 is now available to ChatGPT Pro users and developers globally, with plans to roll it out to additional user tiers in the coming weeks.

The model builds on its predecessor, GPT-4, by offering broader knowledge, improved understanding of user intent, and higher emotional intelligence (EQ). These advancements make it particularly useful for tasks such as writing assistance, programming, and solving practical problems. Developers can also access GPT-4.5 via APIs for applications that require advanced creativity, multi-step task automation, or enhanced conversational capabilities.

GPT-4.5 introduces innovations in training methods, including scalable techniques derived from smaller models, which improve its steerability and nuanced understanding of human input. While it does not currently support multimodal features like voice or video input, it offers functionalities like file and image uploads and advanced planning capabilities for complex workflows.

OpenAI has emphasized safety in this release by employing rigorous testing and combining traditional supervised fine-tuning with reinforcement learning from human feedback. Despite its higher computational demands and cost compared to GPT-4, OpenAI is evaluating its long-term availability based on user feedback.

This release highlights OpenAI's commitment to advancing AI capabilities while addressing challenges such as factual accuracy and usability in real-world applications. By scaling unsupervised learning, GPT-4.5 sets a new benchmark for general-purpose AI models and invites users to explore its potential across diverse domains.

Source