- Published on
Was pizza invented by Americans to sell nylons?
- Authors
- Name
- Peter Hartree
- @peterhartree
Question:
I asked GPT-5 Thinking, Claude Opus 4.1 (Extended thinking), Gemini 2.5 Pro and Grok 4 (Expert).
Grok got the reference. GPT-5 got there with a hint. Claude didn't, and seemed annoyed. Gemini hallucinated a quote from a quiz show.1
Grok has a better sense of humour than other models. Once you get to "what is he referencing?", the answer is just a web search away. And Grok and ChatGPT are much better at web search than Claude and Gemini. Claude was great at saying "I don't know". While Gemini... wasn't.
Footnotes
I sent the prompt to each model three times. Grok's first response "got it" in 2/3 tries; GPT-5 only 1/3. Grok always got it after the "any idea what I was referencing?" follow up. GPT-5 always required that question plus the "focus on the nylon" hint. Claude never got it. And Gemini hallucinated every time. ↩