OpenAI has released a new model, which is not as big as expected, however, it is more like a human! How would we know that they have released this new model?

Thanks to Editor's Pick from our platform, I see the news summary like that by just clicking on it.

But of course I wanted to dig deeper, let’s visit the OpenAI’s platform and check the information that they gave to use first.

SimpleQA

This metric measures how AI answers tricky questions. 

As you can see, the new model, GPT 4.5, has a lower rate of hallucination and the accuracy among all GPT models.

More Human

From the chart above, I understand that OpenAI is scaling up their models. However, instead of just making them bigger, they are trying to make them a little less robotic.

As you can see from the screenshot above, OpenAI suggests that this new model is more human-like! It understands how you feel and uses conversational tone as you see above.

How can you test it?

Of course, if you are reading this, you want to test it, but it is only available through the playground, so let’s see.

They suggest that it is more like a human, so we should test scenarios, humans will face, like renting issues!?

Issue with the Landlord

You're probably renting if you’re not wealthy enough to buy a house. And unless you’re in the top 1%, your landlord is just looking out for their own interests.

So let’s use this prompt.

I have an issue with my landlord, and my contract will end in a month. Can you help me write a polite message to send him before I leave?

GPT 4.5

Here, you can see GPT-4.5’s response. It shows empathy, as they claim, while also ensuring my rights. It includes warm regards and placeholders for me to fill in. I could actually use it next month — after translating it into Turkish, of course!

To see what has changed, let’s see GPT 4o’s answer.

GPT 4o

As you can see, it has noticeably less emotion and empathy. From the message it generates afterward, you can see that it maintains a polite and professional tone.

Final Thoughts

Yes, this model has better features than others, but did we expect this? Honestly, after reading a news article that explains this new model, I was excited, but after reviewing it, I feel like something is missing.

But what is the reason why I felt disappointed? Was it due to the fast-paced AI world increasing my expectations? I don’t know, but I have good news for you guys.

We have launched our new platform, LearnAIWithMe! Here, you can use our assistants without a ChatGPT subscription, tackle data projects, and access daily AI news summaries, which are available to subscribers. See you there!

“Machine learning is the last invention that humanity will ever need to make.” Nick Bostrom