GPT-4.5 Just Dropped (And It’s All About EQ, Not IQ)

OpenAI has officially released GPT-4.5.

The company describes the new model as their “largest and best model for chat yet.” It offers natural conversations, higher emotional intelligence, and a reduction in AI hallucinations.

Unlike other advanced models that focus on complex reasoning (like OpenAI’s o1 or o3), GPT-4.5 doesn’t “think” more deeply—at least not in the traditional sense of chain-of-thought reasoning.

Instead, it scales up the raw power of the GPT approach, pulling in more data and more compute to create a smarter, more empathetic chat experience.

So, what does GPT-4.5 mean for you?

To get the scoop, I talked to Marketing AI Institute founder and CEO Paul Roetzer on Episode 138 of The Artificial Intelligence Show.

GPT-4.5: What You Need to Know

First, there seem to be plenty of benefits to using GPT-4.5.

OpenAI says it's more factually accurate than previous models across a broad range of knowledge. They also say it has a better grasp of when to respond briefly and when to dive deep—a kind of social awareness that plays out in more personable conversations.

(It’s the first model that “really feels like chatting with a thoughtful person,” says CEO Sam Altman.)

But, there are some downsides.

GPT-4.5 is currently only available to the highest-tier ChatGPT Pro users, priced at around $200 a month, as of writing. OpenAI plans to give access to Plus users soon though. It also doesn't yet have Advanced Voice Mode and currently lacks video and screen-sharing.

And, the model itself might not seem earth-shattering to most people. The average user probably won't be blown away by what GPT-4.5 can do. Not unless you're testing an "arsenal" of use cases or prompts every time a model comes out, says Roetzer.

"I think it's more a sign of what's coming versus being some obvious leap forward in capabilities and performance," he says.

With the upcoming release of GPT-5, OpenAI plans to merge GPT models like this one with its reasoning models into one unified system.

However, that doesn't mean we should gloss over the release of GPT-4.5. In fact, says Roetzer, quite the opposite. It's gives us a very powerful clue as to where AI is headed.

Let's Get Emotional

Former OpenAI researcher and cofounder Andrej Karpathy points out a key thing to know about GPT-4.5:

It has significantly improved EQ, or emotional intelligence.

"I think this is very, very fundamental to understand where these things go," says Roetzer.

As models develop more EQ, not just IQ, they become more immediately useful and competent at skills, traits, and jobs that we perceive to be uniquely human. They start to actually understand humans better. Which, in turn, gives them the ability to better support complex tasks in business and life—and changes how we interact with these models.

"It enables interactions that start to feel way more natural," says Roetzer. "It gives AI the feeling of empathy."

The upsides to increased EQ like we're seeing in GPT-4.5 are huge:

• Smoother interactions. If AI can sense you’re upset or struggling, it can pivot its tone and offer more supportive or reassuring suggestions—almost like talking to a good friend or colleague.
• Fewer hallucinations. According to OpenAI’s own data, GPT-4.5 is notably less likely to invent facts. Part of that is likely due to its broader knowledge base. But a higher EQ might also help it pick up on the intent behind questions and avoid random tangents.
• Better collaboration. In everything from customer service chats to creative brainstorming sessions, it could be a major advantage to have an AI that “reads the room” (even if the “room” is just a chat window).

But having models with more EQ is about to make things much more complicated, says Roetzer.

Where It Gets Complicated

With increased EQ, AI models also get much better at persuasion and even manipulation.

"I think this starts to become a truly concerning area of development," says Roetzer.

He notes that these models can already be very good at persuasion. But often these types of capabilities are "red-teamed" out of them, meaning humans put in guardrails and safeguards that prevent models from fully exercising persuasion skills.

(Imagine AI that isn’t just logically convincing, but also empathetic enough to nudge your emotions in subtle ways—or even manipulate you without you realizing it's happening.)

GPT-4.5 is already a significantly more emotionally intelligent system. Which means GPT-5 is likely to have even higher EQ.

And that's what we need to start getting ready for. Because businesses are already behind on using and integrating the raw intelligence displayed by today's models. Add serious EQ into the mix? You're going to catch a lot of firms unprepared, says Roetzer.

Because even if 4.5 itself doesn’t blow your mind, it’s a major stepping stone toward GPT-5—a release that could combine emotional intelligence with powerful reasoning, changing the entire conversation around what AI can (and should) do.

"I think it's probably just very important that people don't gloss over this release like it's the same old thing," he says. "That's not the point. The point is to prepare us for GPT-5."