OPENAI launches GPT-4.5, an AI with more human interactions
An advance towards more natural conversations
GPT-4.5 is distinguished by its ability to recognize social nuances and to respond more intuitively. According to Sam Altman, CEO of Openai, this model is the first to give the impression of converse with a thoughtful person. This improvement is attributed to training combining new supervision techniques and traditional methods, reducing “hallucinations” and refining the personality of the model.
Mixed performance according to the fields
Although GPT-4.5 excels in linguistic understanding and offers improved computational efficiency, it remains below specialized models such as O1 or O3-Mini in certain tests, especially in mathematics and science. This distinction highlights the importance of choosing the suitable model according to specific needs.
A high cost for advanced technology
Innovation has a price. Sam Altman has recognized that GPT-4.5 is a “huge and expensive” model, both in terms of training and operating. For the moment, access is limited to chatgpt pro subscribers, with an extension provided for other users once the GPU capacity has been increased.
Future prospects: towards GPT-5
Openai does not stop there. The company plans to launch GPT-5 by the end of May, incorporating the new O3 reasoning model. This next step aims to unify the different series of models, thus simplifying the user experience and bringing OPENAI closer to its objective of general artificial intelligence.
