Vision-only

Vision-only refers to autonomous-vehicle perception using only cameras. No [lidar](/glossary/lidar), no radar in current architectures. The defining technical posture of Tesla's Full Self-Driving and Autopilot stacks since the "Tesla Vision" transition in 2021–2022, which removed radar from the Tesla Model 3 and Y.

The distinction matters because vision-only is the most cost-aggressive AV architecture and the technical foundation of Tesla's claim that [robotaxi](/glossary/robotaxi) unit economics are achievable at scale. The skeptical case: vision-only fails in failure modes (low sun, rain, occlusion) where lidar succeeds, and human-equivalent vision plus AI inference may not match human vision plus human cognition. Tesla's bet is that scale of training data eventually closes the perception gap; the safety record so far hasn't conclusively confirmed or refuted that bet.