Openai has just presented its new model of artificial intelligence GPT-4.5 which, according to the Chatgpt creator, is its biggest and best cat model, to date. It is important to note that, unlike O1 or O3-Mini, the GPT-4.5 model is not a model of reasoning. It is rather suitable for the current uses of the generative AI and is presented as an improved version of GPT-4O, which is the model currently offered by default on Chatgpt.
GPT-4.5 is a research preversion, which is currently offered to users of the Chatgpt Pro subscription (the Subscription to 200 dollars). But as of next week, Chatgpt Plus and Team subscribers will also be able to take advantage of this novelty. “We share GPT-4.5 in preview in order to better understand its strengths and limits. We continue to explore its capacities and we are impatient to see how people use it in a way that we would not have expected ”Indicates Openai, in its ad.
In any case, thanks to the different techniques used by Openai to improve its AI, Interactions with GPT-4.5 should seem more natural. OPENAI also evokes a better knowledge base, a better ability to recognize the intentions of the user, as well as a greater “emotional intelligence”. Therefore, the new model would be ideal for writing tasks, programming, as well as solving practical problems. On the other hand, Openai discusses a reduction in the number of hallucinations.
An AI that understands you better?
The most interesting improvement could be the ability of GPT-4.5 to collaborate with humans. “The GPT-4.5 understands better what humans mean and interpret the subtle clues or implicit expectations with more nuance and“ QE ”. The GPT-4.5 also shows an aesthetic intuition and stronger creativity. He excels in writing and design aid ”Indicates Openai, on this subject.
OPENAI indicates that to make natural conversations and to improve the understanding of nuances by AI, it has developed a technique that makes it possible to cause larger models with data from smaller intelligence models. To assess improvements, Openai has conducted comparative assessments between GPT-4.5 and GPT-4O with human testers. During this evaluation, the new model surpassed the old for requests for daily life, professional requests, as well as for creative intelligence.
Openai is already preparing the rest
In any case, while GPT-4.5 has just been published, Openai is already working on the next model: GPT-5. This should be more interesting, because, according to a previous publication by Sam Altman, GPT-5 will combine the advantages of models like GPT-4.5 or GPT-4O with those of reasoning models, such as O1 or O3-Mini.
The idea is to simplify the user experience, by offering an AI that automatically adapts to the needs of the user. The latter would therefore no longer have to choose between two types of artificial intelligence according to the task he performs. Currently, for tasks linked to science, mathematics, etc., we can get better results using the O1 reasoning model, which takes a longer time to reflect.
- OPENAI presents GPT-4.5, an improved version of GPT-4O
- This is a research preversion that will be deployed in Chatgpt Plus subscribers and Chatgpt Team from next week
- OPENAI indicates that this is its best model for the cat
- GPT-4.5 has a better knowledge base and has been designed to better understand user’s intentions