GPT-6 Astra drives a real Toyota Corolla through a cone course; Claude Fable 5.1 reaches 45%

GPT-6 Astra has gained the ability to drive a car

DrivingBench gave GPT-6 Astra, Claude Fable 5.1, Grok 4.6, and GPT-5.6 Sol direct control of a real Toyota Corolla's steering, accelerator, and brakes on a cone course. GPT-6 Astra completed the course on its second attempt in 5:22, costing $7.74 in API fees. Claude Fable 5.1 peaked at 45% progress; Grok 4.6 and GPT-5.6 Sol never exceeded 11%. Each model got three attempts inside one continuous chat. The post doesn't disclose total course length, cone spacing, or whether a safety driver intervened. I'd hold off on the '100%' claim until the trajectory replay shows smooth driving vs. constant correction.

Why it matters: Real-car driving test for GPT-6 Astra with a fixed course, head-to-head comparisons, and concrete time/cost numbers. Hits all three HKR axes. Not a sim — actual hardware — which makes it more shareable than most benchmark papers. Score not higher because only the project page ...