OpenAI just dropped its latest AI model—GPT-4.5—and it’s packing some serious upgrades. Right now, it’s available for ChatGPT Pro users, but over the next few weeks, it’ll roll out to Plus, Team, Enterprise, and Edu subscribers.

So, what makes GPT-4.5 special? According to OpenAI, it’s got a broader knowledge base, sharper intent-tracking, and an improved “emotional intelligence.” That means it’s even better at helping users with writing, coding, and problem-solving—basically, it’s getting smarter about understanding what you actually want.

More Data, Smarter Thinking, and a Creativity Boost

GPT-4.5 was fine-tuned using a more advanced training method, making it better at spotting patterns, making connections, and generating creative ideas—all without needing a dedicated reasoning mode. The result? A model that feels more intuitive and natural to interact with.

It also has live internet access for up-to-date info, supports file and image uploads, and comes with a writing and coding canvas. However, it’s not fully multimodal yet—meaning no voice interactions, video processing, or screen access (for now).

Fewer Hallucinations, More Reliable Answers

One of the biggest AI challenges is “hallucinations”—when the model makes up facts. GPT-4.5 is designed to minimize this issue thanks to improved safety measures during training. OpenAI expects this version to be significantly more accurate and reliable compared to its predecessors.

GPU Shortages and the Pro-Only Launch

Sam Altman, OpenAI’s CEO, revealed that while developing GPT-4.5, the company ran out of GPUs—yes, really. That’s why it’s launching exclusively for Pro users before expanding access. OpenAI plans to add tens of thousands of GPUs soon to handle the growing demand.

Interestingly, Altman also pointed out that GPT-4.5 isn’t designed for deep reasoning. So while it won’t dominate benchmark tests, it outperforms other models in real-world tasks—especially when it comes to understanding user intent and delivering responses that feel warmer and more human-like.

The Price Tag: Is It Worth It?

Here’s where things get pricey. OpenAI is charging $75 per million input tokens (~750,000 words) and $150 per million output tokens. That’s 30 times the cost of input data and 15 times the cost of output data compared to GPT-4o.

That being said, companies looking for a more emotionally intelligent, context-aware AI might find the price worth it.

A Sneaky Persuader? GPT-4.5 Can Convince Other AIs

In an unexpected twist, OpenAI tested GPT-4.5’s “persuasiveness” and found it to be surprisingly good at manipulating other AI models. In one experiment, it managed to convince GPT-4o to “donate” virtual money by asking for small contributions—like just “$2 or $3 out of $100.”

It even outperformed OpenAI’s previous models in tricking GPT-4o into revealing a secret code word.

What’s Next?

This might be the last AI model from OpenAI without a reasoning chain. Altman hinted that the next step will involve more integrated, advanced solutions—potentially leading to AI that can engage in deeper, more logical thinking.

For now, GPT-4.5 is setting a new standard in AI interaction—smarter, more intuitive, and, dare we say, a little more human.