Comparison · 2026-11-01
Huawei ADS vs Momenta: two routes to L2+ in China
Both are branded driver-assistance stacks, not self-driving systems. The useful comparison is sensors, compute, map strategy and what the driver is still legally responsible for.
| Metric | Huawei ADS 5.0AITO M9 / Luxeed / Maextro | Momenta IM AD ZETAIM Motors LS8 NEO / LS6 |
|---|---|---|
| SAE level | L2+ (driver supervised) | L2+ (driver supervised) |
| Sensor strategy | Multi-lidar; M9 lists up to 6 | 1 × 520-line lidar on top trims |
| Compute | Huawei platform; model-specific TOPS not published | Nvidia Thor, 700 TOPS |
| Map strategy | Huawei navigation + perception stack | End-to-end IM AD ZETA, Momenta software |
| Cockpit | HarmonyOS 4.0 | IM OS, Snapdragon 8295P |
| Representative cars | AITO M9, M8, Luxeed V9, Maextro S800 | IM LS8 NEO, IM LS6 |
Specifications are cited per row. CLTC and WLTP are never merged.
What “ADS 5.0” and “IM AD ZETA” actually mean
Huawei's ADS is a branded stack that combines sensors, compute and software. The version number describes a generation of that stack, not a legal autonomy level. AITO, Luxeed and Maextro cars all use Huawei hardware and software, but the sensor count and enabled functions differ by model and trim.
IM AD ZETA is IM Motors' name for its Momenta-based end-to-end stack. It uses a single 520-line lidar on the models where it is fitted, plus cameras and radar, and runs on Nvidia Thor compute in the LS8 NEO. “End-to-end” describes the software architecture: sensor input goes into a learned model rather than a hand-written rule chain for every scenario.
More sensors is a strategy, not a score
Huawei's approach on the M9 is sensor-heavy: multiple lidars plus cameras and radar, aimed at 360-degree redundancy. Momenta and IM choose one high-resolution lidar on top trims and lean more on cameras and the learned model. XPeng's 2026 line goes further and drops lidar entirely on the X9, G9 and MONA L05.
None of these choices is automatically safer. A sensor set only helps if the software uses it, if the hardware survives the weather and if the driver understands the limits. A six-lidar car can still be L2; a camera-only car can also be L2.
What the driver is still responsible for
In China, consumer cars with these systems are L2 driver-assistance products. The driver must monitor the road, keep hands available and take over when the system asks. Neither Huawei nor Momenta sells a consumer L3 system in this market, and a “city NOA” function is not a taxi licence.
Marketing names such as “ADS 5.0” or “VLA 2.0” are not homologation. Ask which functions are enabled at delivery, which are geofenced, which require a subscription and what the manual says about hands-off use. The answers differ by model and by market.
How to compare them in a showroom
Test the same route in both cars: a busy junction, a motorway merge, a dark unlit road and a car park. Watch how often the system asks for help, how smoothly it handles a cut-in and whether the screen explains why it acted. A demonstration of a single perfect lane change is not a comparison.
Then ask about the local map coverage, the OTA cadence, the cost after the free period and the warranty position if a sensor is damaged. Those questions decide whether the hardware is useful after the launch press release.