OpenAI unveiled its largest, but not its most advanced, GPT-4.5 model
OpenAI announced the launch of GPT-4.5, its newest and largest artificial intelligence language model. This model, available in preview for Pro users, is positioned as the most "aware" model to date.
Here's What We Know
GPT-4.5 offers improved writing capabilities, deeper knowledge of the world, and a refined personality that makes interacting with it more natural. The model is better at recognising patterns and making connections, making it particularly useful for tasks involving writing, programming and solving practical problems.
However, OpenAI emphasises that GPT-4.5 is not an advanced model and does not offer significant innovations over previous models such as o1 or o3-mini. Nevertheless, it demonstrates higher computational efficiency, reducing errors and "hallucinations".
The model was trained using novel control methods and traditional approaches such as supervised fine-tuning and reinforcement learning based on human feedback. These methods are also being used for GPT-4o.
OpenAI also notes that GPT-4.5 will be an important step towards more functional models that may be labelled as general artificial intelligence in the future.
GPT-4.5 will become available to a wider audience, including Plus, Team, Enterprise and Edu users, in the coming weeks. It is also integrated into Microsoft's Azure AI Foundry platform, expanding its availability to developers and researchers.
Source: OpenAI
