Only three companies in China currently meet the operational standard for Level 4 (L4) autonomous driving, according to Tony Xu Han, founder and chief executive of WeRide, underscoring the challenges facing firms seeking to commercialise robotaxi services.
Speaking at the MEET 2026 Smart Future Conference, Han said the companies that meet the threshold are WeRide, Baidu’s Apollo, and Pony.ai. He outlined a strict definition for L4 operations, arguing that they require fleets of at least 20 to 30 vehicles operating continuously for six months without safety drivers, while generating stable commercial orders.
See also: China Approves First Level 3 Autonomous Driving Permits For Passenger Cars
“In my view, models that still rely on safety drivers, use remote fallback systems, or simply rebrand L2++ as robotaxis cannot be considered true L4,” Han said.
WeRide has deployed more than 1,600 autonomous vehicles globally, including over 750 robotaxis, Han said, representing a roughly 30% increase in fleet size since the company’s U.S. initial public offering a year ago.
Rivals have also reported growing scale. Baidu Apollo said in November that its robotaxi services were handling about 250,000 orders per week. Pony.ai reported that its robotaxi fleet reached 961 vehicles in the third quarter, with an average of 23 orders per vehicle per day in Guangzhou.
See also: WeRide, Uber Launch Public Robotaxi Rides in Dubai via Uber App
Outside China, Han pointed to Waymo as a benchmark for L4 operations, noting that its fleet in North America has grown to about 2,500 vehicles, with significantly higher utilisation rates.
Han also commented on Tesla’s robotaxi ambitions, comparing the U.S. automaker’s approach with Waymo’s. He said Tesla could face difficulties reaching comparable operational performance if it continues to rely on L2 production vehicles such as the Model 3 and Model Y and does not introduce lidar technology within the next few years.
WeRide’s strategy extends beyond robotaxis. Han said the company is developing both L4 robotaxi systems and L2+ production-grade assisted driving, with the aim of reusing core technologies across different vehicle platforms. He added that advances in computing power are narrowing the hardware gap between L2 and L4 autonomy.
See also: Xpeng CEO Compares Tesla FSD With VLA System, Eyes 2026 Catch-Up
The company’s end-to-end autonomous driving solution is already in mass production, Han said, citing collaboration with Bosch on Chery’s Exeed Sterra model.
Reflecting on the downturn in autonomous driving investment around 2019, Han said the industry at that time lacked consensus on both technology pathways and business models. He argued that the focus has since shifted.
“The question today is no longer whether a car can drive,” Han said. “It is about how far it can drive and how safely.”