Okay, I’m always trying to figure out which AI model is actually the best for my projects, and it feels like the answer changes every week. It’s a real headache!
Then I stumbled upon this incredible video where an AI professional put the top four models through a brutal, head-to-head competition. I’m talking a 10-round stress test to see who really comes out on top. The mind behind it tested GPT-5 (Thinking model), Gemini Pro, Grok (Expert Mode), and Claude Opus 4.1.
This wasn’t just a simple Q&A. The creator threw everything at them: coding, reasoning, hallucination tests, and complex problem-solving. It’s one of the most thorough comparisons I’ve seen.
✨ The Most Surprising Results
I was blown away by some of the outcomes. Here are a few highlights from the tests this innovator ran:
- 💻 Website Creation: The expert asked each AI to build a modern, interactive website right in the chat window. Claude was the undisputed champion here, delivering a beautiful and functional site. The others had issues with broken links, bad UI, or choosing random, irrelevant tools.
- 🧠 Reasoning & Vision: This was a tough one. The creator uploaded an image of a pyramid and asked the AIs to identify the correct top-down view. Only GPT and Grok got it right! Gemini and Claude completely failed, which was super interesting to see.
- 🤥 Hallucination Check: To test for made-up facts, the person who shared it asked about the 19th U.S. President’s pet parrot. The trick? The president never had one! I was so impressed that all four models correctly identified the trick and didn’t invent a parrot.
- 📊 Business Projections: This is where things got messy. The YouTuber asked for a 24-month revenue projection but intentionally left out the number of new customers. Instead of asking for that key detail, most models just made up numbers, making the results useless. The expert noted that Claude came the closest to a decent answer but still had to make assumptions.
- 🧩 Animated Maze Solving: In another cool visual test, the creator had the AIs generate a maze, solve it, and animate the solution. While most passed, Claude’s version was visually superior and it consistently created more complex mazes for itself to solve.
🏆 The Final Verdict
After 10 intense rounds, the final tally was… a TIE!
Both GPT-5 and Grok came out on top. However, the creator pointed out that Claude was just one point behind and absolutely dominated any task involving coding or visual generation. It really shows that the best AI depends entirely on what you’re trying to do.
One last fun fact: at the end, the expert had each AI score the competition based on the video’s transcript. Hilariously, every model ranked itself as the winner except for Gemini!
This is just a quick peek. For the full deep-dive and to see every single test in action, you have to watch the original video from the creator!