Author | De Xin Editor | Wang Bo
L3 National Standard Countdown Begins
Currently, the mandatory national standard for L3 is entering its final countdown, and the revised draft of the ADS indicator light regulations is open for public comment. Numerous related signals indicate that L3 autonomous driving may arrive next year.
The market already features vehicle models with pre-embedded hardware foundations for L3-level redundant architectures. Models represented by the Qijing GX7 are poised to become benchmark vehicles for the popularization of L3 autonomous driving architectures. Before L3 commercialization, is it worth for ordinary users to pay in advance?
National Standard Countdown,L3 Autonomous Driving Enters the Eve of Commercialization
At the recently concluded2026 World Intelligent Connected Vehicles Conference, several industry leaders previewed L3 autonomous driving.
GB 44721-2026 "Safety Requirements for Automated Driving Systems of Intelligent Connected Vehicles" will be mandatorily implemented on July 1, 2027, serving as the regulatory safety threshold for domestic L3 compliance.
Industry norms are also laying the groundwork for liability allocation in L3 autonomous driving.
In October, the draft for comments of GB 4785 "Installation Requirements for External Lighting and Light Signaling Devices of Motor Vehicles and Trailers" was published. The document focuses on regulating ADS indicator lights (i.e., the blue lights for intelligent driving). Two core logics are established: ADS indicator lights are optional only for vehicles at L3 level or above; installation and illumination are prohibited for L2 and below. L3 autonomous driving activation automatically turns on the light, and deactivation automatically turns it off, with no manual switching allowed.
Both points indicate thatthe blue light will be a key supporting feature distinguishing L3 autonomous driving liability attribution.
Under the guidance of regulatory authorities, L3 access approval is proceeding in an orderly manner. HiEV has learned that L3 autonomous driving rollout will likely adopt targeted access and limited ODD (Operational Design Domain), prioritizing highways and expressways first, followed by scenario-specific, gradual expansion potentially covering urban areas.
Regulations, lighting indicators, and access mechanisms collectively build the regulatory foundation for L3 commercialization.
Currently, multiple vehicle models on the market have pre-embedded architectures designed for L3 autonomous driving, covering price ranges from over 200,000 CNY to one million CNY.
Huawei's Qiankun Intelligent Driving is the primary technical support behind these models. Before the implementation of L3 regulations, some models collaborated with Huawei Qiankun, adopting pre-embedded L3 autonomous driving architectures based on a hardware foundation featuring eight redundancies (power supply, steering, braking, computing power, perception, positioning, communication, and interaction) to reserve capabilities for subsequent compliant commercialization.
By September this year,the penetration rate of L2 and above assisted driving in China reached 70%.
Jin Yuzhi, Senior Vice President of Huawei and CEO of Yinwang Intelligent Technology, judges that the industry will accelerate its transition from assisted driving to autonomous driving.
L3 vs L2:From Functional Upgrade to Liability Shift,Hardware Pre-embedding is Key
The fundamental difference between L2 and L3 lies in the shift of safety liability.
At the 2026 World Intelligent Connected Vehicles Conference, Jin Yuzhi revealed that autonomous driving imposes higher safety requirements. In the assisted driving phase, if an accident occurs, the driver remains the responsible party. However, at L3 and above, the responsibility shifts to the system.
In other words, after L3 functions are compliantly opened, within limited applicable scenarios, when the system is active and no takeover instruction is issued, accidents caused by autonomous driving system failures will theoretically be compensated by liability insurance provided by automakers and intelligent driving suppliers, rather than borne by the vehicle owner.
According to public data, Huawei Qiankun Intelligent Driving ADS has demonstrated safety levels 5.75 times higher than the average human driving in China over the past 12 months, and this is still based on L2 assisted driving statistics.
L3 autonomous driving will see an exponential increase in safety compared to L2 within limited scenarios.
Furthermore, regarding collaboration mechanisms, L2 requires the driver to act as a fallback, ensuring readiness to take over control at any time when issues arise. In contrast, under L3 conditional automated driving, if the system encounters problems or exceeds the ODD range, it initiates a takeover request to the driver and maintains a takeover gap of at least10 seconds.
A major safety bottleneck in current L2 assisted driving accidents is "human distraction (when takeover is needed)." The L3 architecture utilizes the MRM (Minimum Risk Maneuver) system combined with full-chain redundancy to mitigate the risk of human distraction.
Therefore,enhanced safety in limited scenarios is the first major value proposition of L3.
With improved safety, L3 unlocks new scenarios for users.
Within the ODD, drivers can disengage hands and eyes, allowing them to drink water when thirsty, relax when tired, or even chat on WeChat, watch dramas, play games, or take photos inside the car. Although sleeping is still not permitted, this significantly frees up entertainment and office productivity options. This is the second major value proposition of L3.
Third, L3 may confer exclusive status associated with high-level functions.
The ongoing changes in automotive lighting regulations imply that the blue light will serve as an exclusive identifier for L3 autonomous driving.
Although blue lights are currently common on L2 vehicles, it is foreseeable that they will be phased out on L2 vehicles as new regulations take effect. Next year, blue lights on highway sections will become the exclusive indicator for activating high-level functions.
Given these advantages of L3 over L2, can current L2 vehicles upgrade to L3?
The answer is no, unless the L3 autonomous driving architecture is pre-embedded.
Many consumers mistakenly view L3 simply as a functional upgrade from L2. This is a clear misconception. In reality, having a LiDAR does not guarantee upgrade eligibility; nor does enabling Urban NOA ensure an L3 upgrade.
LiDAR represents only perception capability. L3 requires full-chain redundancy. Its complete hardware architecture includes redundancies across eight modules: power supply, steering, braking, computing power, perception, positioning, communication, and interaction. This ensures that in the worst-case scenario, if one set fails, another can immediately compensate.
At the software level, manufacturers are also required to establish fallback and traceability capabilities, including MRM scenario libraries, takeover capability monitoring, safety archives, and comprehensive safety management systems.
In short, L3 capabilities rely on natively pre-embedded hardware foundations.Ordinary L2 vehicles without pre-embedded full-chain redundant L3 architectures cannot upgrade to L3 via OTA.
From the perspective of automakers and technology suppliers, the hardware architecture pre-embedded for L3 represents only visible costs; liability and compliance costs constitute a significant portion as well.
To some extent, these software and hardware costs are a major reason why some intelligent driving suppliers and automakers choose to bet on L2 and L4, skipping L3. Essentially, manufacturers are unwilling to bear the legal, claims, engineering, and commercial burdens attached to L3.
However, Huawei Qiankun believes that L3, as the critical boundary between assisted and autonomous driving, cannot be skipped and represents a necessary cycle for liability verification.
Huawei Qiankun's decision to deeply participate in national standard formulation, collaborate with automaker partners, treat L3 as a mandatory step, promote native full-chain redundant L3 architectures, and provide joint safety fallbacks with automakers reflects the differences in business models and risk tolerance between Huawei Qiankun, as a primary technology supplier behind L3, and other vendors.
How to Choose in 2026: Is L2 Enough, or Pay in Advance for L3 Architecture?
Although most L2 assisted driving models currently on the market cannot upgrade to L3 due to hardware architecture limitations, L3 is not far away. Mainstream vehicles priced under 300,000 CNY now offer options with pre-embedded L3 autonomous driving architectures. For instance, the Qijing GX7, launched in September, starts at a limited-time benefit price of 229,900 CNY, with the full series equipped with a full-chain redundant architecture designed for L3.
This vehicle alone features four LiDARs (one forward-facing 896-line dual-path image-grade LiDAR and three solid-state blind-spot-filling LiDARs), totaling 38 high-performance sensors across the vehicle. More critically, the Qijing GX7 incorporates redundancies across eight systems: power supply, steering, braking, communication, positioning, computing power, perception, and interaction. This is the underlying logic distinguishing it from ordinary L2 architecture vehicles.
Rough estimates suggest that the hardware increment brought by L3 autonomous driving architecture—such as four LiDARs, redundant steering, dual braking, dual power supplies, and dual domain controllers—costs approximately 30,000 to 40,000 CNY. Yet, the starting price of the Qijing GX7 remains in the 200,000+ CNY range, even offering larger dimensions than same-class five-seat SUVs. This reflects GAC's determination, in partnership with Huawei Qiankun, to create a benchmark for L3 popularization.
Previously, many believed that vehicles with L3 autonomous driving architectures would be concentrated in high-end models or produced in small batches for trial purposes. However, the Qijing GX7 proves that family-oriented SUVs targeting the mass market can also carry this solution, demonstrating that the L3 full-chain redundant architecture is not exclusively for luxury vehicles costing over one million CNY.
The Qijing GX7 is poised to represent the inclusive deployment of L3 autonomous driving architectures. The core factor lies in GAC partnering with Huawei Qiankun to design the vehicle from the initial development stage according to L3-oriented full-chain redundant architecture standards.
For GAC, this architectural foundation is key to breaking free from homogeneous competition in the 200,000-300,000 CNY segment and building differentiated competitiveness. For Huawei Qiankun, the Qijing GX7 serves as a benchmark template for implementing its L3 technology roadmap.
Although L3 functions are not yet commercially available, pre-embedded L3 hardware brings experience value-add before then, such as more mature and stable L2 assisted driving experiences:
Multi-layer redundancy designs and additional sensors enhance L2 safety capabilities from perspectives like perception distance and fault response. For example, the 896-line dual-path image-grade LiDAR can stably identify low obstacles as small as 14 cm at a distance of 120 meters, thereby improving safety against small targets in highway scenarios.
While both offer urban/highway assisted driving functions, this system's "fallback capability" is stronger than that of ordinary L2 vehicles.
This is the dimension of short-term experience.
From the long-term perspective of vehicle resale value, many worry that their smart driving capabilities will become obsolete after using a car for two or three years. The design goal of the L3 autonomous driving architecture is oriented toward high-level autonomous driving. Over the vehicle's approximate 10-year usage lifecycle, subsequent OTAs can continuously unlock higher-level capabilities.
In the future used-car market, values may gradually diverge based on architectural capabilities: one category comprising vehicles with pre-embedded L3 architectures possessing upgrade potential, and another consisting solely of L2 architectures unsupported for L3 upgrades.
Just as cars without assisted driving are hard to sell today,L3 autonomous driving architecture capabilities are expected to become an important value benchmark for new and used cars in the future.
L3 autonomous driving will open up new imaginations for future mobility and is the inevitable path toward L4.
When L4 is realized, as Jin Yuzhi depicted at the World Intelligent Connected Vehicles Conference, "mobility will be safer and more convenient; future cars will park themselves, charge themselves, wash themselves, drop off children, and buy coffee autonomously."
Starting from L2 assisted driving, Huawei Qiankun is gradually building this future mobility model from a holistic commercial scenario perspective. Recently, Huawei Qiankun signed a battery swap agreement with CATL, aiming to integrate the chain from intelligent driving to battery swapping for energy replenishment. In future autonomous/assisted driving states, vehicles will proactively drive to swap stations for recharging.
The Qijing GX7's full-chain redundant architecture oriented toward L3 serves as the foundation for inheriting this future mobility paradigm.
If you plan to keep your vehicle for more than 5 years and care about its long-term intelligent residual value, consider prioritizing models equipped with L3 architectures within the same budget. This represents better assisted driving experiences and longer intelligent lifecycles. Moreover, technological upgrades will liberate valuable driving time.