OpenAI’s new GPT-4.5 model is a better, more natural conversationalist

When you are already busy in the last few days New model releasesOpenai connects the week to one with one Research review of GPT-4.5. The company is still wearing the new system as the biggest and best model for chat. In an early test, Openai said that people are a more natural conversation, which has a more natural conversation with GPT-4.5, to deliver warmth and show some kind of emotional intelligence.

In an instance shared by Openai, a person says a test is a difficult time after a test failed. The company’s previous models, including GPT-4O and O3-MiniBefore submitting a long list of unwanted recommendations, you can contact the individual, GPT-4.5 is a different courtesy. “Do you want to talk about what happened or do you just need a distraction?

The winning by GPT-4.5 is the result of Openai progress in uncontrolled learning. With an inexperienced learning, a machine learning algorithm is given an unmarked data set and is released to their devices to find examples and concepts. The GPT-4.5 company “does not think” as the most modern substantial models, but the new model has developed Openai architectural accessories and gained more information and power accounts. “The result is a model, which is a more extensive knowledge and a model that causes reduced hallucinations, which is a deeper understanding of the world,” he says.

Speaking of reduced hallucinations, Openai was better measured in GPT-4.5 in this regard. While passing SymbolDue to the ability to respond to large language models “flat but difficult knowledge questions”, a clear design that tests large language models, “GPT-4.5”, GPT-4O and even a hallucination rate beat with 37.1 percent. Obviously, the new model does not solve the problem of AI hallucinations in general, but is one step in the right direction.

Despite the relative powerful aspects of GPT-4O and O3-Mini, GPT-4.5, it is not directly replaced for models. Compared to Openai justification systems, GPT-4.5 is a more general purpose model. ” In addition, it means it does not like a local multimodal, but does not work with features like this Sound modeVideo or screenshot. It is also a “very large and calculated intensive model.”

It is good to think like a stone that steps on the systems he planned to offer GPT-4.5 in the future. In fact, Sam Altman said so much before this month shared the company’s road mapTo celebrate in GPT-4.5, “Our last chainless well-thought-out model” – referring to the fact that the new system does not solve problems by resolving the problems step by step by step-by-step by Openai. His successor, GPP-5, likely to combine many Openai’s latest technologies, including Border O3 model. Openai has repeated today that GPT-4.5 has plans to bring the EQ’s EQ for future models, “the EQ, the EQ,” EQ, “EQ,” EQ. “

Meanwhile, Pupiling for Subscribers can start using GPT-4.5, today and the team has a slot to get access to the next week with users.