OpenAI Unveils GPT-4.5: A Leap Towards More Human-Like AI Interaction:

OpenAI has released a new AI language model, GPT-4.5, which is said to be the most sophisticated to date. It is currently available in research preview for all ChatGPT Pro users and Global Developers.

The new model of ChatGPT comes shortly after the emergence of rival models such as DeepSeek-R1 that have forced companies such as OpenAI to innovate at a faster pace.

OpenAI's CEO, Sam Altman, proclaimed via his page on social media platform X that this model is the first to approach the feeling of conversing with an actual sentient being and that it feels like talking to a thoughtful person. In his words, “I have had pretty good advice coming from an AI on several occasions which has surprise me."

From what Altman stated, OpenAI has put a lot of investment into this model as they are using the terminology “giant, expensive” as a description of it. The CEO of OpenAI continues to state that this model will not “crush” the benchmarks like other models. He explains, “It's a different kind of intelligence and there's a magic to it i haven't felt before.”

What does GPT-4.5 bring?

As per IDRsolutions, OpenAI claims that initial tests show that users experience GPT-4.5 as more natural and easier to use than the earlier versions. The model also includes more knowledge, better alignment with user intents, and greater emotional intelligence which make it particularly good at writing, programming, and other practical tasks. In addition, OpenAI says that GPT-4.5 will be able to get rid of hallucinations, which is when AI provides false or illogical responses.

Nevertheless, proficients AI will not consider GPT-4.5 as a frontier AI model. OpenAI also points out that the model allows users to interact with the system more efficiently, and it is computationally more efficient, but it does not provide new features compared to models o1 or o3-mini, to be precise.

GPT-4.5 was trained with new supervision strategies on top of traditional supervised fine-tuning and reinforcement learning from human feedback. This has allowed reaching the goals of reducing hallucinations, providing emotionally aware interactions, and improving its performance in collaborative tasks.

GPT-4.5: Availability

To start off, ChatGPT Pro users as well as developers have access to GPT-4.5. It is set to be available to Plus and Team users in a week, followed by Enterprise and Education users. The model is now also part Microsoft’s Azure AI Foundry platform which widens availability to their enterprise clients.