Guiyang 2026: WeRide Beats Huawei ADS — The Smart-Driving Gap UK Buyers Must Understand

The second edition of China's citywide ADAS competition chose Guiyang's old town districts — judges described it as the toughest course in 30 rounds of testing, harder even than the Tianjin event held earlier this year. Twelve cars entered. The signature challenge was a narrow, two-way single-lane street requiring a three-point turn: the only test on the day with a pass rate below ten per cent. Eleven of the twelve cars failed it. The twelfth was the GAC Aion N60, powered by WeRide's autonomous driving system. It completed the manoeuvre unaided, accumulating zero takeovers across the entire course, and took first place with a score of 117.81 out of 120 — the highest total of the series so far. WeRide also powered the third-placed car, the Xingji Yuandian ET (scored 96.5). NIO's ES8, armed with three LiDARs and more than 1,000 TOPS of compute, came second on 97.5. It completed most challenges well but its human driver ultimately had to take over at the three-point turn.

Nine days before the Guiyang event, Huawei formally announced ADS Pro V5.0 — the update was prominently billed as bringing campus NCA (Navigate on Autopilot) for more than a million parking locations, emergency auto-braking improvements, and, notably, three-point turn assist. The timing makes the Guiyang results particularly instructive. All four Huawei ADS-equipped cars — the Zhijie R7 (ADS Max V4.1.5), AITO M9 (ADS Max V5.0.0 — the newest firmware), Avatr 06T (ADS Max V4.1.5), and Denza N7 (which runs Momenta's V2.5.2 stack, not Huawei ADS, but is widely grouped with Huawei-ecosystem cars) — failed the three-point turn test. The M9, on the fresh V5.0 firmware that supposedly adds the exact capability being tested, drove into the oncoming lane and required human intervention. The R7 performed strongly everywhere else — D1EV noted it handled an unexpected blocked road with zero hesitation — making the single-point failure more striking rather than less. This is not necessarily evidence that Huawei ADS is poor overall. It is evidence that a single announced feature and a controlled real-world test are different things, and the gap between them can be wide.

WeRide's win with 200 TOPS and a single LiDAR has been framed by some Chinese media as proof that raw compute and sensor count don't determine ADAS capability. That is partially true and worth unpacking. The WeRide system on the Aion N60 is purpose-built for urban robotaxi and mobility service environments — dense traffic, intersections, stop-start movement at low speeds — which is precisely the competency being tested at Guiyang. The same stack won third place on the Xingji Yuandian ET, confirming it was not a fluke. What the result does not prove is that 200 TOPS is sufficient for highway NOA at 120 km/h in variable weather, or that a single LiDAR handles all European driving scenarios. The NIO ES8's higher hardware investment did produce a strong result across most of the course — it came second, not last. Huawei's strongest car, the R7, showed real capability in unscripted moments. The key lesson is narrower: for urban low-speed manoeuvring, a well-trained model on moderate hardware can beat a poorly-calibrated model on premium hardware. Software maturity in the specific task domain is the differentiator, not the TOPS number in the brochure.

None of the cars in the Guiyang competition are currently on sale in the UK. But several of the brands are — Jaecoo and Omoda are Chery subsidiaries; Leapmotor sells the C10 and B10 here; and if Aion's Ray lineup eventually reaches right-hand-drive markets, these competition results will matter. More immediately, UK buyers evaluating Chinese EVs encounter smart-driving claims constantly in marketing material. A car's ADAS supplier (WeRide, Huawei Qiankun, Momenta, Bosch) and compute figures (TOPS) are increasingly prominent in brand communications. This competition is a useful calibration: the gap between announced features and tested performance can be wide and narrow quickly; a headline announcement nine days before a public test is not the same as deployed capability. For UK buyers, the practical takeaway is straightforward: smart-driving features available in China are typically not active on UK export versions, due to geofencing and regulatory differences — the UK BYD Seal carries no LiDAR, the UK Leapmotor C10 has 84 kW DC charging because it runs an older battery architecture, and assisted driving in UK trim is L2 lane-keep and adaptive cruise, not city NOA. The China competition results are relevant as brand intelligence — they tell you something about the quality and trajectory of a manufacturer's software team — but they are not a spec you will experience behind the wheel in Britain.

The Guiyang result is a good data point — genuinely useful intelligence — but only if you hold it correctly. It tells us that WeRide's urban ADAS stack is competition-class right now; that Huawei's ADS V5.0 launch, while impressive in breadth, had a gap between announced capability and real-world execution within nine days of announcement; that sheer hardware does not compensate for software calibration in specific scenarios; and that China's self-driving ecosystem is developing fast enough that the competitive landscape changes on a monthly basis. What the result does not tell us is which stack will prevail over three to five years, or which of these systems will eventually reach UK roads. For UK buyers today: the cars you can actually buy in Britain carry L2 ADAS regardless of what their Chinese siblings do in competition. That gap should narrow — Leapmotor has a roadmap, BYD's God's Eye system is rolling out in China — but timelines are uncertain and the regulatory path for anything beyond L2 in the UK is long. Check the UK spec sheet, not the China competition podium, when you are deciding whether to order.