These Researchers Made AI Drive a Toyota Corolla to Get In-N-Out
Only one LLM successfully drove a real car to get In-N-Out. Here's which one — and why the others failed.
Why it matters
Three researchers tested GPT, Claude, and Grok on real-world autonomous driving tasks using a Toyota Corolla. Only one model could reliably execute the multi-step workflow (navigation, ordering, driving). The result illustrates the gap between LLM capability in controlled benchmarks and performance in embodied, real-world environments with safety stakes.
The key facts
11 to knowThree models tested: GPT, Claude, Grok
Real car: Toyota Corolla
Task: autonomous drive to In-N-Out (multi-step: routing, ordering, navigation)
Success rate: one model successful, two failed
Source: Wired, Oct 7 2026
No specific failure modes or performance metrics disclosed
Three researchers tested GPT, Claude, and Grok controlling a real Toyota Corolla
Only one model successfully completed the task (autonomous drive + In-N-Out ordering)
Two models failed in ways that matter for autonomous-system deployment
Test published October 2026
Models tested in real-world conditions, not simulation
Go to the source
Wired AIwired.com
Publisher excerpt: Three engineers put GPT, Claude, and Grok in charge of a real car. Only one of them was successful.