Summary: A new report just compiled a list of the smartest AI models that are available to the public. The list includes models from Google, OpenAI, DeepSeek, xAI, and Anthropic.
At this point, it doesn’t really know which AI model is smarter, as most of them are extremely intelligent. So, pitting them together is like splitting hairs for the most part. However, it’s still interesting to know which models can do the most. Companies are on the fast track to achieving AGI, so we want to know which one is the closest.
A new report from The Neuron took the top-performing AI models from companies like Google, OpenAI, Anthropic, xAI, and DeepSeek and had them face off. The top AI model probably won’t surprise you, and there’s also a standout model.
What’s the smartest AI model on the market right now?
The models used in this test are o4-mini, o3-mini, o1-mini, Gemini 2.5 Pro Preview, Grok 3 mini Researching, DeepSeek R1, and Anthropic Claude 3.7 Sonnet Thinking. These are pretty much the most powerful models on the market. They were run through the Artificial Analysis Intelligence Index, and each one was given a specific numerical score.
One thing to note is that benchmark scores aren’t the only way to gauge a model’s intelligence. However, they do provide some solid information about where they stand. Looking at the scores, a few things pop out.
Firstly, we see that OpenAI’s most powerful model, o4-mini, was able to beat out the competition by getting 70 points. That was followed closely by Gemini 2.5 Pro Preview, with 68 points. So, based on the score, these two models are pretty much neck-and-neck.
The next few models only trail by a few points. What’s surprising is that Grok 3 mini is so high. xAI is relatively new compared to the other companies on the list. It hasn’t been training models as long as OpenAI or Google, so it’s surprising that it’s able to follow so closely behind the competition.
The list trails off with DeepSeek R1 (60), Claude 3.7 Sonnet Thinking (57), and o1-mini (54). DeepSeek, just a few months ago, broke the internet with its model that pretty much came out of nowhere. Now, it looks like it has fallen behind the competition. OpenAI and Google have released some new and more powerful models in the meantime.
Price
As pointed out by the report, the best overall model is o4-mini, but Gemini 2.5 Pro Preview was the best when it came to the price. This is something that might push more developers to use Gemini’s API. Both models are comparable, so developers will be able to implement more AI into their workflow for less.
King for how long?
It’s important to know that the world of AI models constantly changes. Companies deliver new features and updates to models at a moment’s notice. OpenAI’s model is at the top of the dog pile, but it’s not by much. There’s no telling if a new model from Google, DeepSeek, or Meta will come around and dethrone it.
Meta recently announced that its most powerful model is on its way. We don’t know if that model will give all of the top AI companies a run for their money.
