Pure visual solutions for advanced driving? Competition in China’s auto market intensifies. Car manufacturers release new models at a rapid pace. In the last two months, around 50 new energy vehicles hit the market. The domestic new energy vehicle market faces a surge of new releases. These new energy cars offer surprising prices. The Leidao L60 attracts the most attention. Under the Baas battery rental model, this mid-sized electric SUV costs only 149,900 yuan (21060$). Buyers pay a one-time price, followed by a monthly battery rental fee. This pricing strategy helps the new model gain popularity quickly. NIO announced that the Leidao L60 has “exploded in orders.” Before the Leidao L60, another domestic new energy vehicle also claimed “explosive sales.” Xpeng’s second brand, MONA, launched its first model, the M03, priced between 119,800 and 155,800 yuan (21890$). Xpeng reported that the MONA M03 secured over 10,000 pre-orders in 52 minutes and surpassed 30,000 in 48 hours. To ensure timely delivery of the MONA M03, Xpeng announced two production expansions within ten days.
The two popular new cars target different markets and have significant price differences. Recently, they often get compared to illustrate a point: the high-level intelligent driving market is continuously evolving. Car companies are shifting their technical routes for intelligent driving once again, bringing pure vision solutions back into the spotlight. This trend aligns with the industry’s current price competition, as pure vision solutions theoretically cost less. The pure vision camp is making strides. Jiyue is one of the first domestic companies to abandon lidar solutions in favor of pure vision. Backed by automotive and tech giants Geely and Baidu, Jiyue has clear advantages in funding and technology. Its first model transitioned from an early lidar solution to a finalized pure vision approach. Jiyue views pure vision as a highly challenging yet rewarding autonomous driving technology. CEO Xia Yiping stated that, in the long run, pure vision offers advantages in system costs, iteration speed, data loops, and even end-to-end development. Thus, Jiyue boldly decided to abandon the lidar plan. Jiyue became the second global company and the first domestic company to adopt a pure vision solution. However, it’s important to note that when Jiyue’s first model, Jiyue 01, launched in late October last year, the domestic intelligent driving field still favored lidar solutions. Mainstream new energy companies viewed lidar as essential to compete with Tesla’s FSD high-level intelligent driving system. In less than a year, the trend has completely shifted. The data from the Gaishi Automotive Research Institute shows that multiple brands have launched pure vision solutions. Beyond Extreme’s ASD solution, Huawei introduced ADS SE, Zhuoyue launched the Chengxing platform, and NIO’s Lido released OSD. Xiaopeng Motors will soon launch the AI Eagle Eye vision solution. GAC Group plans to mass-produce GARCIA in 2026. Suppliers like Jianzhi Robotics, SenseTime, and Wenyan Zhixing also offer pure vision solutions. A new “pure vision landscape” is emerging. Currently, Huawei’s basic version of the pure vision driving solution, ADS SE, has debuted on the new Wanjie M7 PRO and Deep Blue L07. It will also appear on Deep Blue S07 and Zhijie S7 PRO. NIO’s pure vision route has been applied to Lido L60. Zhuoyue’s Chengxing platform includes models like Baojun Cloud and Baojun Cloud Sea, with Baojun Cloud Sea already featuring end-to-end capabilities. Xiaopeng’s MONA M03 is another model with a pure vision solution. The top version includes the Xiaopeng XNGP system, powered by two NVIDIA Drive Orin chips and dual cameras, enabling urban assisted driving. The upcoming Xiaopeng P7+ will officially feature the new AI Eagle Eye vision solution. Gaishi Automotive Research Institute believes that domestic players in smart driving are trending towards pure vision solutions. The core driving force is the continuous innovation of large models and end-to-end algorithm technology. Additionally, the increasingly competitive market compels everyone to choose more cost-effective driving solutions.
The advantages and disadvantages of pure vision are clear. Lower costs stand out as an obvious benefit. Take the Baojun Yunhai as an example. Some argue that its standard memory driving function does not qualify as a “proper” urban NOA. However, with a starting price of 100,000 yuan (14050$), many see it as a typical representative of high-level intelligent driving functions reaching the mass market. Yet, the 100,000 yuan (14050$) standard for high-level intelligent driving did not quickly open the market for Yunhai. The low-end image of the Baojun brand remains a significant barrier to its products and market growth. Additionally, the hardware capability of only 32 TOPS raises concerns about the car’s true intelligent driving performance. Still, the Yunhai allows us to directly feel the current push for high-level intelligent driving. According to sales data from Gaishi Auto, from January to August 2024, domestic models with standard high-level intelligent driving functions sold nearly one million units, a 147.9% increase compared to the same period last year. We can predict that as lower-cost pure vision solutions ramp up production, the domestic high-level intelligent driving market will likely continue to grow at a faster pace. Returning to the analysis of pure vision advantages, Gaishi Auto Research Institute notes that the core strength of pure vision is its low dependence on sensors. Compared to heavy lidar and millimeter-wave radar solutions, it significantly reduces overall costs. This approach offers better cost-effectiveness for achieving intelligent driving functions. The pure vision solution has significant drawbacks. It heavily relies on continuous algorithm iteration and vast amounts of high-quality data. It also demands high computational power and infrastructure from supercomputing centers. These factors create substantial hidden costs. Previous data showed that weekly iterations and training for a pure vision solution cost between 20 to 30 million yuan. Many automakers and smart driving suppliers choose pure vision solutions. They aim to balance high and low-end markets. For example, Huawei distinguishes its smart driving offerings into basic and advanced versions. The basic version follows a pure vision route, leveraging low costs to penetrate the market. The advanced version adds lidar to maintain a presence in the high-end market. NIO’s situation is similar. The brands Leidao and NIO adopt different approaches: pure vision and lidar, respectively. This strategy creates clear brand differentiation in smart driving. Leidao helps NIO lower its market entry point to 200,000 yuan (28100$). Importantly, in competition with external rivals, Leidao’s first model, the L60, holds its ground. Compared to Tesla’s Model Y, which also uses pure vision, the L60’s starting price is 43,000 yuan (6040$) lower. Against competitors with lidar, like Zeekr 7X and Xpeng G6, the L60 remains competitive in smart driving. The Gai Shi Automotive Research Institute believes the L60 has advantages over competitors. It offers a lower purchase threshold and a more complete charging system. Additionally, it shows no significant weaknesses in smart driving and cabin features. The new car launched and orders surged. According to Li Bin, founder, chairman, and CEO of NIO, the market response is very enthusiastic. Orders exceed expectations. The main task now is to quickly increase production capacity.
The market potential for pure visual routes is significant. Some in the industry argue that pure visual routes can seamlessly connect with the popular end-to-end concept. The technology foundation of end-to-end architecture enables AI to think like humans. Pure visual routes aim to assist or replace human vision with cameras. They mimic human sight through algorithms. The underlying technologies are closely linked. However, some voices in the industry raise doubts. The concept of pure visual routes lacks clear, unified standards. There is no authoritative definition of what constitutes a true pure visual route. Currently, Tesla leads the way. It relies solely on pure vision, without using lidar, millimeter-wave radar, or ultrasonic radar. Many domestic car companies, however, still rely on additional sensors. They equate not using lidar with being purely visual, yet they still use millimeter-wave and ultrasonic radars. From a technological maturity perspective, it is understandable that domestic companies are slower. Tesla began focusing on pure visual solutions in 2021. Most domestic brands have only started to transition in the last couple of years. Developing mature visual algorithms will take time. The real concern for the future is whether pure visual routes can support higher levels of intelligent driving. After surpassing L2, can pure visual routes function without the precision of lidar? The entire industry must collaborate to find answers.