The Superbestest AI Is? Yes, Grok
May 7, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
Visualize a pleasant day on April 7, 1945. The largest and most expensive battleship was attacked by several hundred aircraft. The ship suffered direct hits from bombs and torpedoes. Imagine that you are pushing spotting boats around the deck so the blast from the Yamato’s guns would not damage the boats. Lots of action. Lots of effort. Lots of money. The Yamato capsized and its forward munitions magazine exploded. Hasta la vista, Yamato.
I thought of the good ship when I read “Everyone’s Switching from ChatGPT to Claude — But New Tests Say Neither Is the Smartest Free AI, and the Real Winner Might Surprise You.” The test was not comprehensive: No Gemini, China China China, or Mistral.
An outfit named OmniCalculator did the “research.” Let me cut to the finish line. Grok 4.2 is the “smartest AI around.” Ah, you don’t believe it? Why is Grok the big dog in being “smart”?
The write up says:
When it comes to the quantifiable math ability of these AI chatbots, the smartest free AI model is, rather surprisingly, Grok. xAI’s Grok 4.2 model specifically. That doesn’t mean anything about its writing style and ability, or anything else chatbots can do, but it does suggest that it might have the edge in math prowess.
Okay, but I thought the Google technology was “good” in math. I think I heard the PR wanna be from DeepMind make that clear to dinobabies like myself. The OmniCalculator created a chart, and it showed how good the Grok 4.2 system was in this three hombre race.
Here is the chart of totally reproducible research (like most modern research these days):
© Omnicalculator, 2026
Okay, so if I input one of those videos YouTube thinks I crave along with news about auto shop scams, French slang, and ancient lost technology, what’s my result.
Here’s the quiz from Math Queen: Solve for x when x = x / 5. You can watch the MQ entity method on the new Boob Tube.
Claude said:
x = x / 5 is only true when x = 0.
ChatGPT said:
x = 0
Grok said:
Final Answer: x=0
The dinobaby (me) looked at the expression and said, “Zippo.” The smart software and I presented the correct answer.
Grok may be better than ChatGPT if these reported data are valid. I am not sure Claude and ChatGPT are significantly better than one another.
But Grok? Yes, Grok is the winner. The cited article says:
[Grok] is far less likely to backtrack or alter its conclusions mid-process. That’s great for reasoning and logic, but not much help in mimicking the smooth tones that make other models feel more polished.
Yes, a win for Elon. ChatGPT and Claude are smarmy and that’s why people like them.
My view is different. The basic “idea” for the BAIT (big AI tech firms) smart software is Google’s Attention Is All You Need paper. Therefore, fiddling with the knobs and dials of the algorithms produces variances. But if the only material one has to build a house is cardboard. Houses built of that material share a fundamental characteristic: Yep, cardboard. How long do some of those township and favela dwellings last?
I think Omnicalculator did a good job of presenting Grok as a winner. Mr. Musk’s testimony at the trial contesting the status of OpenAI did not strike me as the output of a Sam Snead or Annie Oakley. Therefore, I conclude that this excellent bit of research is content marketing designed to present Grok as the direct talking, nuts-and-bolts smart software everyone needs.
I wonder why DeepMind and the China China China models were not in the competition. Oh, I know. It’s like youth sports. Pay to play, right?
Now back to the biggest and most expensive battleship. Took hits. Sank. End of story for the effort put into moving stuff around the deck.
Stephen E Arnold, May 7, 2026
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
