What improvements offers and how to test the latest version of artificial intelligence

This version is not a Reasoning Model As O3, Deepseek R1 or Grok-3, capable of reviewing their “thoughts” before providing a definitive response, but the figures shared by the company are quite promising.

In fact, according to OpenAI, GPT-4.5 has a “hallucinations” index minor compared to GPT-4O, O1 and O3-mini. This means that the model Invent less information and, therefore, Lads less.

Those responsible for Chatgn promise greater precision in issues of general knowledge compared to previous models, stating that this new model “It has a broader and more deep knowledge of the world”.

In search of similarities and coincidences, their makers compared this new version with O1, conceived to spend more time to “think” before responding. These models are optimized to improve precision and coherence In your answers, reducing the amount of errors.

The model has a better understanding about what humans consult him and interprets subtle signs or implicit expectations with greater nuance and emotional intelligence.

The conclusion is that, sometimes O1 can provide more information, the way in which GPT-4.5 “flows” by offering your answers is much more advanced. Which implies that inherent reasoning and power have reached a new value scale.

During the presentation they have dedicated a important part to security. And while the main objective is that AI achieves greater precision in the answers, they also pretend that the information is safe.

For now, GPT-4.5 It will be a paid. The first who can start using it are the subscribers of Chatgpt for the company’s 200 dollars plan. But in the next few days it will begin to be available in the Chatgpt Plus Subscription.

Normally OpenAi always leaves their best models for those who pay, so it is logical to think that it will take a few months to launch a Free version and that it will have some type of limit.

Sam Altman the OpenAi CEO shared a fairly extensive and detailed post in X, in which it announced the official arrival of Chatgpt-4.5 and offered some considerations about the new version of its chatbot.

Altman said that chatgpt-4.5 is “The first model that gives the feeling of talking to a reflexive person” And that he had several moments in those sitting in a chair while using the program and was surprised “upon receiving good advice from an AI.”

In this breakdown he recognized the Weakest point of this new model. “It is a giant and expensive model,” said the businessman. “We have grown a lot and we have run out of Gpus,” Altman said, while in Openai’s press release it is detailed that the firm had to resort to Microsoft Ai Supercomputers from Microsoft for training.

He also confirmed that they will add “tens of thousands of GPUs next week” and that they will also launch the new version so that the users of the more plan.

These statements arise after the shake that caused DeepSeek Chinese artificial intelligence created by Liang Wenfeng, whose startup already has the accelerated launch of its new model R3initially planned for May 2025.