As China Gears Up for Level 3 Autonomous Driving, the Rest of the World Remains Focused on Level 2++

Recent iterations of China’s, and by some measures, the world’s leading auto show have been forums to demonstrate the cutting edge of automotive technology, and the growing gap between faster-moving and more vertically-integrated Chinese Original Equipment Manufacturers (OEMs) and their more legacy competitors. Perhaps the most obvious takeaway from Auto China 2026 is that the gap between Chinese OEMs and the rest of world remains, making it increasingly difficult to discern any real gap between the offerings of increasingly homogenous Chinese automakers. Large displays, Artificial Intelligence (AI) agents and immersive multi-media have become par for the course at a time when Chinese OEMs desperately need competition avenues beyond pricing in a congested market that cannot sustain the number of incumbents.

This is where Level 3 autonomous driving has entered as an emerging differentiator, positioning eyes-off operation on highways and the addressable time for rich digital experiences that unlocks as a way for premium OEMs to set themselves apart from the competition. Earlier (2025) testing permits have since been followed by new liability frameworks and safety standards setting out minimum risk maneuver requirements, providing a regulatory environment for highway eyes-off systems to take off from 2027. In contrast to many areas of automotive technology, highway eyes-off technology is one area in which Chinese OEMs are following, rather than leading; moreover, they are following Western OEMs that have adopted and subsequently abandoned Level 3 in favor of Level 2++.

Over the course of a few days in February 2026, Mercedes-Benz and BMW both independently announced that their pioneering Level 3 autonomous systems would be removed from their flagship S Class and 7 Series models, with both instead pivoting toward next-generation, eyes-on Level 2++ systems. While both of the pioneering Level 3 systems emphasized eyes-off operation in very specific contexts, these subsequent Level 2++ systems favor eyes-on assistance in the form of point-to-point hands-off driving.

Regulatory accommodation for Level 3 is highly restricted outside of China, so Level 3 systems became a poor mixture of low consumer value, high OEM risk exposure, and high system prices (US$7,000+). Both BMW and Mercedes-Benz are premium OEMs that benefit from being first to market with almost any experience, but it is telling that, having achieved the brand-boost associated with pioneering a new technology, both have opted to proceed with Level 2++, which offers a far more compelling mixture of high consumer value, low OEM risk exposure, and much more modest system pricing.

Tesla and GM are both enjoying take rates in the 30% to 40% range for their own Level 2+ offerings, but BMW and Mercedes-Benz can both expect much better adoption rates for their Level 2++ systems, and the prospect of genuine returns on their Autonomous Vehicle (AV) offerings. For example, Mercedes-Benz has priced its MB. DRIVE ASSIST PRO system at US$3,950 for a 3-year period, and assuming system Bill of Materials (BOM) costs in the region of US$500 to US$600 at scale, a 35% take rate would see the OEM generating a return of around US$800 per vehicle. Even more encouraging news for Mercedes-Benz is the emerging evidence from GM's OnStar that premium vehicle consumers typically opt to renew their Level 2+ subscriptions once the initial period has lapsed.

Of course, if Level 3 becomes established as the norm for premium vehicles in China, then both Mercedes-Benz and BMW must find a way to quickly develop a competitive Level 3 system to regain market share in China. The ideal option is a scalable approach that adds redundancy in the form of tertiary sensors, additional compute, and High-Definition (HD) map layers to replace the supervisory role of the driver in certain ODDs, unlocking eyes-off operation.

Even in a more favorable regulatory environment with more opportunities for consumers to benefit from eyes-off operation, the HMI of Level 3 remains a challenge. The handover process, which should feel natural to the consumer, requesting driver operation in a timely and predictable fashion that does not shake consumer confidence in the system, will be an area of competitive differentiation between OEMs.

At the same time, it is important for OEMs to bear in mind that Level 2++ will increasingly come with its own HMI challenges. Even without additional redundancy, the use of transformers and End-to-End (E2E) AI approaches is resulting in a comfortable, humanized hands-free experience that can address driving tasks from point-to-point with only very irregular driver intervention. Passively observing a near-perfect driving experience over a protracted time period is a task for which humans are ill suited, and questions remain over how realistic it is to expect a human driver to be ready to intervene after hours of uninterrupted automation. A potential solution is the use of reasoning models to self-describe what the system is seeing, its intentions, and the reasoning behind a change in course, leveraging active narration to keep drivers more robustly in the loop.