Tesla FSD vs Waymo vs Mobileye vs Baidu Apollo
Side-by-side technology comparison of the leading autonomous driving systems — sensor stacks, SAE autonomy levels, geographic availability, pricing models, and engineering approach.
| Category | Tesla FSD | Waymo One | Mobileye SuperVision | Baidu Apollo Go |
|---|---|---|---|---|
| SAE Level | Level 2+ (driver must supervise) | Level 4 (no driver in ODD) | Level 2+ (driver must supervise) | Level 4 (select geofenced cities) |
| Sensor stack | 8 cameras only (no LiDAR, no radar on newer models) | LiDAR + 29 cameras + 6 radars | 11 cameras (EyeQ chips), no LiDAR | LiDAR + cameras + radar + ultrasonic |
| AI approach | End-to-end neural network (V12+): pixels → controls, trained on fleet video | Modular: perception, prediction, planning pipelines; HD maps required | Hybrid: neural networks + rules; REM crowdsourced HD maps | Modular + learning; Apollo platform open-source core |
| HD maps required | No (map-free by design in V12) | Yes (essential for operation) | Yes (REM mapping from fleet) | Yes (high-definition maps for ODD) |
| Availability | Consumer Tesla vehicles (US, CA, and expanding internationally) | Robotaxi — Phoenix AZ, San Francisco CA, Los Angeles CA | OEM-licensed to ZEEKR, Polestar; Europe and China | Robotaxi — Beijing, Shenzhen, Wuhan, Chongqing, and more |
| Pricing model | $99/month subscription or one-time purchase (~$8,000 US) | Ride-hailing fares (similar to Uber/Lyft pricing) | Vehicle option fee (OEM-specific); no consumer subscription | Ride-hailing fares; some free trial rides in China |
| Fleet size | Millions of FSD-equipped consumer vehicles | 700+ Jaguar I-PACE and Waymo-spec Zeekr robotaxis | Thousands of OEM-licensed vehicles | 1,000+ robotaxis in commercial operation in China |
| Training data | Billions of real-world miles from Tesla fleet (shadow mode + interventions) | 20+ million real-world miles + Waymo Simulation (Waymax) | REM crowdsourced mapping + simulation | Real-world + Apollo simulation environment |
| Compute platform | Tesla FSD Chip (HW3 / HW4 72 TOPS) | Custom Waymo compute units; Google Cloud TPU for training | Mobileye EyeQ6 chip | NVIDIA Drive + Baidu Kunlun chips |
| Key strength | Scale — millions of vehicles improve the model simultaneously | Safety record — lowest reported disengagement rate in robotaxi operation | OEM integration — camera-only system fits existing vehicle supply chains | China market — largest commercial L4 robotaxi fleet in China |
System profiles
Tesla FSD
The only consumer-grade end-to-end neural autonomous driving system at scale. V12+ uses a single neural network from camera pixels to drive commands, trained on billions of real miles. Available as a subscription to all Tesla owners.
Waymo One
The only commercially operating Level 4 robotaxi in the US. Uses a rich sensor suite (LiDAR, radar, cameras) and HD maps. Achieves the lowest disengagement rate of any deployed AV system. Operates in Phoenix, San Francisco, and Los Angeles.
Mobileye SuperVision
Camera-only L2+ ADAS system licensed to OEMs (ZEEKR, Polestar). Uses EyeQ chips and REM crowdsourced HD maps. Targets the volume OEM market where cost, packaging, and supply chain integration are key constraints.
Baidu Apollo Go
China's largest L4 robotaxi fleet with 1,000+ vehicles operating commercially across multiple cities. Built on the open-source Apollo platform. Targeting full driverless operation in Chinese Tier-1 cities.
Frequently asked questions
Tesla FSD vs Waymo: which is better?
They serve different purposes. Waymo (SAE Level 4) is safer within its operational design domain — no driver needed in select cities. Tesla FSD (SAE Level 2+) requires an attentive driver but works on any road nationwide. Waymo has a better current safety record; Tesla FSD has broader reach and is improving rapidly with fleet data.
Is Tesla FSD fully autonomous?
No. Tesla FSD is SAE Level 2+ — the driver must remain attentive and ready to intervene at all times. Full autonomy (Level 4/5) without a human driver is not yet available on consumer Tesla vehicles as of 2025.
How much does Tesla FSD cost?
Tesla FSD is available as a monthly subscription ($99/month in the US as of 2024) or as a one-time hardware + software purchase option. Pricing varies by country and changes over time. Check Tesla's website for current pricing.
How does Tesla FSD work?
Tesla FSD V12+ uses an end-to-end neural network: 8 cameras feed images into a large neural model that outputs steering, acceleration, and braking commands directly — no hand-coded rules. The model is trained using imitation learning on millions of human driving clips collected from the Tesla fleet. Earlier versions (V11 and below) used a hybrid of neural networks and explicit rule-based C++ code.
Explore the research behind these systems
Browse the academic papers powering FSD, Waymo, and other AV systems: