Hello, ladies and gentlemen. Thank you for standing by for Li Auto's Second Quarter 2025 Earnings Conference C all. At this time, all participants are in listen-only mode. Today's conference call is being recorded. I will now turn the call over to your host, Ms. Janet Zhang, Investor Relations Director of Li Auto. Please go ahead, Janet.
Thank you, Darcy. Good evening and good morning, everyone. Welcome to Li Auto's Second Quarter 2025 Earnings Conference Call. The company's financial and operating results were published in a press release earlier today and are posted on the company's IR website. On today's call, we will have our Chairman and CEO, Mr. Xiang Li, and our CFO, Mr. Johnny Tie Li, to begin with prepared remarks. Our President, Mr. Donghui Ma, and CTO, Mr. Yan Xie, will join for the Q&A discussion. Before we continue, please be reminded that today's discussion will contain forward-looking statements made under the safe harbor provisions of the U.S. Private Securities Litigation Reform Act of 1995. Forward-looking statements involve inherent risks and uncertainties. As such, the company's actual results may be materially different from the views expressed today. Further information regarding risks and uncertainties is included in certain company filings with the U.S.
Securities and Exchange Commission and the Stock Exchange of Hong Kong Limited. The company does not assume any obligation to update any forward-looking statements except as required under applicable law. Please also note that Li Auto's earnings press release and this conference call include discussions of unaudited GAAP financial information as well as unaudited non-GAAP financial measures. Please refer to Li Auto's disclosure documents on the IR section of our website, which contain a reconciliation of the unaudited non-GAAP measures to comparable GAAP measures. Our CFO will start his remarks in Chinese. There will be English translation after he finishes all his remarks. With that, I will now turn the call over to our CEO, Mr. Xiang Li. Please go ahead.
[Foreign language].
Hello everyone, this is Xiang Li. Thank you for joining today's Earnings Conference Call. In the second quarter of 2025, we delivered over 110,000 vehicles, bringing total revenues to RMB 30.2 billion. According to insurance registration data from the China Automotive Technology and Research Center, we captured a 13.4% market share in the RMB 200,000 and above NEV market in China, making yet another quarter where we have top segment sales among Chinese auto brands. As of the end of July, our cumulative deliveries exceeded 1.36 million vehicles. I would like to extend my sincere gratitude to those 1.36 million families who have placed their faith in us, as well as our partners, shareholders, and employees for being part of this journey. We launched Li Mega Home in April this year. The sales results far exceeded our expectations, with approximately 3,000 units sold monthly.
Fourteen months after launch in March last year, Li Mega has become the best-selling MPV priced above RMB 500,000 since May 2025, and then the top seller among all BEVs in the same price range since June 2025. These achievements demonstrate our ability to quickly identify issues and grow, while standing as a powerful testament to our product value as validated by the market and test of time. The refreshed Li L series that we launched in the first half of this year experienced sales fluctuations since June due to sales and service system adjustments and other market factors. Despite these challenges, in June and July, we maintained our top three positions in the RMB 200,000 and above NEV market. We will draw on Li Auto's ability to thrive in challenging conditions and swiftly complete the restructuring of our sales and service system.
This includes enhancing sales and delivery capabilities, building an end-to-end marketing system, and boosting sales team morale. Meanwhile, we will focus on creating value for our users to respond more effectively to market dynamics. We officially launched Li i8 on July 29th, with deliveries starting on August 20th. Li i8 showcases our latest innovations and redefines best SUVs by combining the versatility of off-roaders, the ride and handling of luxury sedans, and the comfort of MPVs. Its pioneering design is a key reason why users choose Li i8, validating our product philosophy. Since its launch, our focus has been on accumulating user word of mouth for this innovative product. We are especially pleased to see a test drive satisfaction rate of over 97%. As customers experience the product firsthand, Li i8 continues to gain momentum.
Since deliveries began on August 20th, we received glowing reviews from the initial batch of customers. We expect cumulative deliveries of Li i8 to exceed 8,000 units by the end of September. We're confident that over time, just like Li Mega, Li i8 will establish itself as the benchmark in its price range. Additionally, preparation for our five-seat BEV SUV, Li i6, is progressing smoothly. We plan to launch Li i6 and commence deliveries at the end of September. Li i6 will expand our product lineup to cater to a broader audience base. To support our BEV launch, we operate the largest charging network among automakers in China. As the automaker with the fastest growing charging network over the past few years, Li Auto strives to empower BEV users with a hassle-free experience. As of now, we have more than 3,100 charging stations equipped with over 17,000 charging stalls.
Among these charging stalls along the highways, over 1,000 supercharging stations cover China's busiest nine east-west and nine south-north highway routes, a 150-kilometer interval on average. Additionally, more than 2,100 urban supercharging stations are spread across over 260 cities, with an average coverage radius of 3.5 kilometers in first and second-tier cities. We're confident that we'll reach our 4,000 station goal by the end of this year. Our charging network also leads the industry in terms of charging rate. Currently, all of our charging stalls support charging rates of 250 kilowatts or higher, with over 61% being 4C and 5C chargers. All of our upcoming charging stations use 4C and 5C charging stalls, further enhancing the ultra-fast charging capability of our energy replenishment networks. Furthermore, we continue to explore and build a new ecosystem for our automotive charging services.
In June, we launched China's first pass-through supercharging station in Changzhou, which functions like a gas station. It allows users to stay in the car while supercharging multiple vehicles at the same time, making the charging experience more efficient and more convenient. We're also testing autonomous charging robots, aiming to redefine smart charging in the future. We remain committed to ongoing investments in research and development to further solidify Li Auto's leading position in intelligence. We believe that intelligence is becoming an increasingly critical, maybe even the most critical driver of users' purchasing decisions. It will also reshape the automotive and mobility industry. We expect our AI investments this year to exceed RMB 6 billion, which will be allocated to infrastructure development as well as product and technology development.
On August 20th, we officially rolled out our proprietary VLA large model driver, and Li Xiaohu is an agent with the delivery of Li i8. As the industry's first VLA architecture to be delivered in production vehicles, the VLA driver large model is scheduled to be deployed on all the AD Max models via OTA updates by mid-September. The VLA large driver large model integrates spatial and linguistic intelligence with behavioral strategy. It boasts enhanced capabilities in precise spatial perception, chain-of-thought reasoning, and decision-making, natural language interaction just like human drives, and routes with diffusion generation and optimal path selection. By adopting a technical architecture that more closely resembles human intelligence, the VLA driver large model significantly enhances user experiences in areas such as defensive driving, smoothness and comfort, three-point turns, continuous task execution, and underground garage navigation.
Furthermore, through ongoing and reinforcement learning, it can become increasingly adept at understanding user preferences. Meanwhile, our proprietary foundation model, Mind GPT-powered Li Xiaohu, has evolved from a voice assistant into an intelligent agent. The agent's enhanced capabilities greatly expand the in-car service ecosystem. For example, it can act as a life assistant to connect with food delivery apps and pick up orders at restaurants, or activate vehicles' external cameras and make payments by scanning and paying fees at parking lots. Its memory capability not only increases the diversity of tools, but also makes Li Xiaohu more emotionally resonant and personalized. Going forward, we will continue to build our AI capabilities and harness our R&D achievements to continuously advance product iteration and evolution. With the vehicle as a platform, we're committed to bringing the physical and the digital world, extending the benefits of AI to every user.
Beyond AI, our R&D also focuses on core areas such as electric drive technology and 5C supercharging. For instance, our proprietary drive motors, including key components like silicon carbide power chips, have significantly boosted vehicle energy efficiency and range. Another key R&D focus is 5C batteries. We have developed in-house battery cells and a thermal management system that can better support ultra-fast charging and long driving range. The 5C battery can maintain a high charging power of over 300 KW while it charges from 0%- 80% of SOC. Additionally, the battery lifespan is exceptional. Even after 1,500 full charge discharge cycles in supercharging mode, battery health remains above 80%, which is industry first. We will continue to leverage our in-house R&D advantages to further bolster product competitiveness and optimize production costs by developing key technologies in-house and manufacturing in-house or through joint ventures.
Turning now to our sales and service network, a recent organizational restructuring will empower us to build our store-centric user operation capability, which we believe will help boost sales conversions and improve how we communicate our brand and product value. Meanwhile, we will continue to define our network structure by expanding and upgrading coverage in major auto parks and shopping malls. We will gradually increase the number of stores in lower-tier cities to reach more user cohorts. To better support our NEV launches, our store expansion schedule has been strategically front-loaded this year. As of now, we operate over 550 retail stores across more than 150 cities nationwide, with net additions of about 50 stores and 700 display spots since the end of last year. Throughout our journey, we have always believed in doing what we believe is right, even when it's difficult.
From EREVs, Li One, and Li L series to BEVs, Li Mega, and Li i8, our commitment to developing advanced technologies and better products remains unshaken, no matter how much adversity we face. This year, we've demonstrated progress in BEV innovation with offerings like Li Mega Home and Li i8. Looking ahead, we will remain devoted to creating superior EREVs and BEVs while strengthening our brand. We'll continue to leverage technological innovation to lead industry transformation. Through our long-term persistent efforts, we will aim to provide our users with cutting-edge technology, extraordinary services, and meaningful companionship. I will now turn the call over to our CFO, Johnny, to walk you through our financial performance.
Thank you, Xiang. Hello, everyone. I will now walk you through our second quarter financial performance. Given time constraints, my remarks today will be limited to some key financial results. All figures will be quoted in RMB unless otherwise stated. For further details, including the corresponding U.S. dollar amounts, we encourage you to refer to our earnings press release. Total revenue in the second quarter was RMB 30.2 billion, down 4.5% year-over-year, and up 16.7% quarter-over-quarter. This included RMB 28.9 billion from vehicle sales, down 4.7% year-over-year, and up 17% quarter-over-quarter. The year-over-year decrease was mainly due to lower average selling price caused by production mix changes, customer interest subsidy, and higher sales incentives, partially offset by increased vehicle delivery. The sequential increase was mainly due to the increase in vehicle deliveries. Cost of sales in the second quarter was RMB 24.2 billion, down 5.2% year-over-year, and up 17.3% quarter-over-quarter.
Gross profit in the second quarter was RMB 6.1 billion, down 1.8% year-over-year, and up 14.1% quarter-over-quarter. Vehicle margin in the second quarter was 19.4%, versus 18.7% in the same period last year, and 19.8% in the prior quarter. The year-over-year increase was attributable to lower average cost of sales, mainly due to cost reduction, partially offset by lower average selling price caused by production mix changes, customer interest subsidies, and higher sales incentives. Vehicle margin remained relatively stable over the prior quarter. Gross margin in the second quarter was 20.1%, versus 19.5% in the same period last year, and 20.5% in the prior quarter. Operating expenses in the second quarter were RMB 5.2 billion, down 8.2% year-over-year, and up 3.8% quarter-over-quarter. R&D expenses in the second quarter were RMB 2.8 billion, down 7.2% year-over-year, and up 11.8% quarter-over-quarter. The year-over-year decrease was primarily due to decreased employee compensation.
The sequential increase was mainly impacted by the pace of new vehicle programs and higher expenses to support our product portfolio expansion and technology advancements. SG&A expenses in the second quarter were RMB 2.7 billion, down 3.5% year-over-year, and up 7.4% quarter-over-quarter. The year-over-year decrease was primarily due to decreased employee compensation, partially offset by increased marketing and promotion activities. The sequential increase was mainly due to increased marketing and promotion activities. Income from operations in the second quarter was RMB 827 million, up 76.7% year-over-year, and 204.4% quarter-over-quarter. Operating margin in the second quarter was 2.7% versus 1.2% in the same period last year and 1% in the prior quarter. Net income in the second quarter was RMB 1.1 billion, down 0.4% year-over-year, and up 69.6% quarter-over-quarter.
Diluted net earnings per ADS attributable to ordinary shareholders were RMB 1.03 in the second quarter, versus RMB 1.05 in the same period last year and RMB 0.62 in the prior quarter. Now, turning to our balance sheet and cash flows, we maintain a robust cash position of RMB 106.9 billion as of June 30th, 2025. Net cash used in operating activities in the second quarter was RMB 3 billion, versus RMB 429.4 million in the same period last year and RMB 1.7 billion in the prior quarter. Free cash flow was negative RMB 3.8 billion in the second quarter, versus negative RMB 1.9 billion in the same period last year and negative RMB 2.5 billion in the prior quarter.
For our business outlook for the third quarter of 2025, the company expects the delivery to be between 90,000 and 95,000 vehicles, and the quarterly total revenues to be between RMB 24.8 billion and RMB 26.2 billion. This business outlook reflects the company's current and preliminary view on its business situation and the market conditions, which is subject to change. That concludes our prepared remarks. I will now turn the call over to the operator to start our Q&A session. Thank you.
Thank you. If you wish to ask a question, please press star one on your telephone and wait for your name to be announced. If you wish to cancel your request, please press star two. If you're on a speaker phone, please pick the handset to ask your question. For the benefit of all participants on today's call, please limit yourself to two questions. If you have additional questions, you can re-enter the queue. If you are a Mandarin speaker, please ask your questions in Chinese first, then follow with English translation. Your first question comes from Tina Hou from Goldman Sachs. Please go ahead.
[Foreign language]. Thanks for taking my question. I have two questions. The first one is regarding our sales volume, because this year we've seen the L series, their sales volume has been declining. How does management plan to achieve our full-year target through product, marketing as well as channel strategies going forward? The second question is regarding our self-developed chips. Can management give us some update on this front? Thanks!
[Foreign language].
Translating for this mock here. There has been a big focus on sales recently, and because our product consists of BEV and EREVs, I will answer this question from two perspectives. First of all, on the EREV front, we will solidify our market position through intelligence. As a leader in range extended technology, we're enhancing product competitiveness through major upgrades in assisted driving. Starting in September, all V80 Max models across our EREV lineup will receive the VLA-assisted driving system. We delivered the VLA Preview version to users alongside VLA delivery. The performance improvement that VLA brings can be compared to the improvements from ChatGPT 3.5 to 4.0. The VLA driver large model has 4 billion parameters, over 10x higher compared to our previous end-to-end model. This is analogous to an increase in brain capacity. This has significantly improved experiences in two core areas.
In daily driving, it delivers substantially improved smoothness and comfort. Many users comment that it's now very difficult to tell apart whether it's the car driving itself or the human driving. In addition, in terms of parking, users have praised our newly introduced VLA summon function and parking functions as it actually addresses pain points while at the same time giving them a better experience. From the industry perspective, we believe the VLA architecture is going to be highly aligned with human intelligence evolution. We're seeing more and more industry participants recognize and join the VLA account. We believe that future competition of assisted driving will hinge on iteration speed, for which reinforcement learning is the key. The simulation environment required for reinforcement learning has been built and is now operational.
The core technology behind this is the world model, which is used to reconstruct and generate the scenarios and data required for reinforcement learning. Moving forward, we will leverage this system to rapidly iterate the VLA model so as to maintain its industry-leading position. That's for EREVs. On the BEV front, this year our lineup with Dennis will be released over time through two major models, gradually opening up new growth opportunities. Li Mega has steadily achieved monthly sales of over 3,000 units, stabilizing our BEV sales foundation. Since its launch, our Li i8 has received very positive feedback from test drives. We're now ramping up production capacity and are striving to deliver between 8,000- 10,000 units cumulatively by the end of September. In September, we will launch Li i6. The model has a very attractive design. It drives very well.
It balances comfort and sportiness and has excellent space experience. It precisely meets the needs of younger consumers and has the potential to become our sales driver in the BEV segment.
[Foreign language].
On the marketing front, we'll be adopting tailored strategies for different markets. Our sales servicing system is now built around one headquarter and 23 provincial regions, with the core principle being localization with strategies tailored to local markets. For example, in the northern region, we will be focusing on promoting EREVs because of their advantages in range and performance in winter. In southern regions, we're prioritizing BEVs, highlighting sales selling points like energy efficiency, spacious interior, and smart features, aligning closely with the local consumer preferences. We'll increase market investment, moving away from our previous assumption of great products will sell themselves. We'll fully embrace digital operations, building a sophisticated digital marketing platform to track and optimize the entire consumer journey, from audience targeting and lead generation to opportunity conversion. This enables precise decision-making on online marketing investments and significantly improves campaign effectiveness.
On the channel side, for tier one, tier two, and tier three cities, we're optimizing our store portfolio to improve consumer acquisition and conversion, with a core focus on optimizing existing resources. We'll optimize store locations. We're proactively relocating stores in suboptimal positions for lower traffic areas to ensure higher consumer reach and operational efficiency. Meanwhile, we'll balance store mix. We're refining the ratio between mall-based stores and central stores in auto parts, leveraging mall store strengths in high consumer traffic areas to attract visitors, while utilizing central store high conversion capability to close sales, creating a seamless acquisition to conversion loop and enhancing overall operational efficiency. To accelerate our coverage in lower-tier cities, we're adopting a lightweight model called Star Plant to deploy stores to unlock new growth potentials.
Compared to traditional stores, these Star Plant stores require lower investments and have a shorter setup cycle, enabling rapid development in core locations within tier four and tier five cities. By increasing store density, we'll boost brand visibility and fully tap into the potential of lower-tier markets and inject fresh momentum into our overall sales growth.
This is Yan. Let me share some information about our in-house design chip. The chip is successfully taped out and returned at the beginning of this year, and it is currently undergoing in-vehicle testing, and everything is in good shape. We expect to deploy it on our flagship models and deliver it to users next year. It takes about three years from setting up the project to its shipment. As far as I know, it is the fastest among similar efforts. The performance is quite satisfactory. Compared to the most powerful chips on the market, it could provide 2x performance when running GPT-like large language models and 3x when running vision models like CNNs. We designed a novel data flow architecture in which model computation is mostly driven by data, not instructions like other architectures.
In this way, the chip could achieve higher parity at the runtime, and we believe it is more suitable for large neural networks. The data-driven logic is orchestrated by our in-house design compiler, allowing the hardware to be more efficient and running at a higher frequency than most comparables in the market. Quite different from other AI chips on the market, we adopted a true hardware-software co-design approach. The chip, the compiler, the runtime system, and the Halo operating system are designed together from the beginning, so we can vertically integrate these hardware and software modules to a more powerful AI inference system more easily, and it could keep on scaling in the future.
With the landing of VLA large models on vehicles, we observed that computing power increase could translate to ADA's performance increase better than before, which means the higher the computing power, the better the performance, and it is more predictable. We have very strong confidence our innovative architecture, as well as the full stack development capability, could become our continuous differentiated capabilities and grow even stronger in the future. Thank you.
Thank you. Your next question comes from Tim Hsiao from Morgan Stanley. Please go ahead.
[Foreign language]. My first question is about the adjustment to the sales system. What's the current progress of the adjustment to the auto sales system? Could you comment on the rationale behind the adjustment and the result you expect to achieve? In the meantime, could these changes adversely affect short-term sales performances and the pace of the new model launches? That's my first question. Thank you.
[Foreign language].
Translating for Mr. Ma here. In August, we went through a major reorganization of our sales team. Now, currently, the headquarters directly manages 32 regions, and we also established new departments for sales and service operations and marketing. We've also restructured the store site selection team, strengthened our training academy, and enhanced our vehicle delivery teams. The new organizational structure is now fully operational and operating smoothly. In driving the sales system transformation, we adopted a four-step approach and internal methodology. Frontline experts clearly are identified as the key stakeholders. We focused on addressing three core needs of the frontline experts while aligning with the company's goal of delivering the best direct sales store experience. The first need we're trying to address is to ensure the frontline experts earn a competitive income. The income of frontline sales experts is directly hinged upon two things: order volume and sales needs.
We have established a dedicated marketing department to generate more online lead needs and restructured the site selection team to boost offline foot traffic to stores. We now give authority for sales strategy development during non-peak periods. We've also strengthened the training academy, equipping our experts with high-quality product message and materials and tools to improve order and conversion rates. Second, we will provide them with growth opportunities. We've built a dual track career advancement path for frontline staff, offering both specialist and then managerial tracks. We've launched targeted training programs for store managers, clearly defining long-term career pathways. This enhances team stability and strengthens frontline execution capability. Thirdly, we will enhance efficiency. We've reduced reporting layers and adopted a flatter and faster decision-making structure. Now, feedback channels, dedicated quality, and operations teams have been established, ensuring frontline concerns are heard, addressed, acknowledged, and acted upon.
This leverages our direct sales model to quickly resolve on-the-ground challenges. The sales system transformation is not a reactive adjustment, but it's a proactive strategic move to better align with frontline needs and further strengthen our direct sales advantage. While there may be some short-term adaptations, in the long run, it will continuously enhance both user experience and team effectiveness. Regarding new vehicle launches, our timeline remains unchanged, and the support is now even stronger. Li Auto's new model rollout is progressing strictly according to our established product roadmap. Li i8 was successfully launched at the end of July, with deliveries beginning on August 20th. Our production facilities are currently operating at full capacity. By the end of September, we expect to cumulatively deliver over 8,000 units of Li i8 and aim to challenge the 10,000 unit mark. Looking ahead, Li i6 will also launch as scheduled in September.
The sales system transformation is designed to provide stronger support for these new launches, optimizing sales and service processes, elevating the consumer purchase experience, and ensuring smooth new model launches and driving rapid sales growth. Thank you.
Thank you. Your next question comes from Yingbo Xu from CITIC. Please go ahead.
[Foreign language]. I have two questions. The first one is that Li i8 has adjusted some of the configurations. What's our future strategy for the product and SKU for future models? My second question is how do we see better the gross margin level under the consumption of the third quarter's sales and revenue guidance? Thank you.
[Foreign language].
First of all, we will be for sure reducing the number of SKUs. Back to the time of V1 and Li L9, we'll be focusing on a single SKU and making sure it's competitive, it's maximized, and providing customers with the best product and value for money. In terms of technological platform and products, we will be iterating faster. The issue we're facing today is by the time we play one hand, our peers have played two hands. The pace of the iteration is equivalent to letting our peers see our cars while we're playing. We must increase the pace of the iteration of both products and technology to make sure that we can play faster. We're confident that by 2026, both in terms of vehicle products and AI, we will be having a stronger lead compared to the time of Li L9 back in 2022. Thank you.
This is Johnny for the third quarter gross margin. I think with the current sales volume and revenue guidance, we expect to maintain our gross margin at about 19%, just as the second quarter and first quarter.
Thank you. Your next question comes from Paul Gong from UBS. Please go ahead.
[Foreign language]. My two questions, the first one is regarding Li i6. We see the allowed scope. Can you please share more regarding the product positioning, the launching timetable, as well as the marketing plan? My second question is regarding your overseas strategy. I'm not sure what is the latest thought, and I would appreciate if you can share more color on this. Thank you.
[Foreign language].
For Li i6, I'm pretty confident that the Li i6 will become a most competitive product in the large five-seater BEV SUV market because it has a unique exterior design, it has industry-leading space and comfort, and a very long range, and one of the best real-world range. It would also be equipped with our industry-leading VLA large model. It would also be the best handling Li Auto product till today. On Li i6, we'll be taking a more user-centric, a new way of marketing. We'll be speaking to our product value and experience in a way that users can understand more easily. For sales, deliveries, and services teams, the training will also be more effective. Lastly, we will seriously take PR defense. We'll make sure that problems are addressed before they're extended.
[Foreign language].
Translating for Mr. Ma here. When our first product, Li One, was delivered in volume, it was back in the beginning of 2020. We look at 2020 and through 2024 as the first stage of development for Li Auto, where we focus on the domestic market and on EREV products. The second phase is from 2025 to 2027. The markets we sell in will be formally expanded to global and domestic markets, and products extended to electric vehicles. The third phase will be after 2027, where we'll be focusing on all four autonomous driving and new form factors of AI agents. 2025 is the initial year of our global strategy. On the R&D front, we have built development centers in Germany and the U.S.
On the sales channel side, we are formally starting to build an after-sales and after-sales organization and building an overseas team, as well as a range of IT systems. On the product front, all of the products that we're developing to be launched in 2026 will be taking into account global regulations. In terms of product roadmap, in terms of marketing go-to-market strategy, we also have solidified our overall plan to focus on Middle East, Central Asia, and Europe.
market is both an opportunity and a challenge. For example, overseas, Li Auto is not a very well-known brand, just to name a few of the challenges. The overseas market expansion will take time and effort, but we have enough patience and resolution. In fact, the overseas market expansion has been one of the top medium to long-term strategies for Li Auto. Thank you.
Thank you. Your next question comes from Xue Deng from CICC. Please go ahead.
[Foreign language]. So I have two questions. The first is, we have seen that our operating cash flow or net cash flow outflows further expand in the second quarter. May you explain what the main reasons and how should we view the subsequent cash flow situation in the future? The second question is about the autonomous driving. We can see some key changes in our autonomous sectors. Also we see the regulatory tightening on autonomous driving. Will that affect our future development and also our integration and also deployment of our new VLA functions?
This is Johnny. I will take the first question. I think in the second quarter, as our payment term by the end of last year was three months to four months, we are paying off most of the payables of last December, November, and January, February, basically. This is why in the second quarter we have some negative operating cash flow. As you may know, we have already adjusted the payment term to 60 days to respond to the regulatory requirements and the industry trends. In the third quarter, the operating cash flow negative, the operating cash flow yield can be will be enlarged in the third quarter. We expect that will improve the cash flow in the fourth quarter if we can have a better sales volume balance in the fourth quarter. Thank you.
[Foreign language].
Translating for Mr. Ma here. To directly answer the question, yes, our assisted driving team has seen a few departures recently. Some level of talent movement is not uncommon in this blossoming industry, and the percentage of people involved is actually very small. We'd like to extend our sincere thanks to those who have left for their contribution at Li Auto. Together, we have truly delivered many great products and experiences for our users. Our current assisted driving team operates with a very clear organizational structure, a well-developed talent pipeline, and strong strategic reserves of talent. In light of recent changes, we have appointed new leaders who bring a more youthful energy to the roles. They carry extensive experience from their past work and offer an international technical perspective. We have the kind of innovative thinking that will drive the team forward.
Li Auto's strong brand appeal and promising prospects continue to attract outstanding talents from across the industry and bring fresh energy to our team. We're steadily advancing our VLA product delivery. The VLA preview version we have already delivered features VLA driving, VLA parking, and VLA summon capabilities. We expect to roll out a complete version with scheduled shipment to all V80 Automax users in September, adding the VLA command functions to deliver a more convenient and intelligent assisted driving experience. We also have another major update scheduled for October to create additional user value. The increased regulatory focus on assisted driving really represents an important step towards healthy and orderly growth for this industry. From our perspective, a stricter regulatory environment helps the truly capable companies with real technical expertise to stand out. At Li Auto, safety and compliance have always been our top priority when developing new technologies.
As we adapt to these evolving standards, we've been strengthening our testing protocols and quality control systems. This not only improves VLA's technological maturity and reduces potential risks, but also helps us to deliver a safer and more reliable assisted driving experience for our users. Throughout VLA's development, we've maintained focus on user value and compliance boundaries, constantly striking a balance between safety validation and experience innovation. We're confident and fully capable of maintaining consistent progress in VLA's development and ensuring its timely rollout. Thank you.
Thank you. Your next question comes from Ming Hsun Lee from Bank of America. Please go ahead.
[Foreign language]. Right now, Li Auto already built a lot of 5C supercharging stations. In the future, will you open your charging station for other brands? How will you operate the charging station if you start to give other brands to use? Will it impact your own brand's car owners' user experience? My second question is regarding the autonomous driving technology. Right now, you have a very good VLA technology. After VLA, what do you expect the development direction of autonomous driving and the technology iteration? Thank you.
[Foreign language].
For translating for Mr. Ma here, for Li Auto vehicle owners, I can assure that the experience with our charging stations will always be the best. I can summarize the experience with four words: coverage, speed, experience, and value for money. First of all, coverage. We're absolutely number one among all automakers in terms of the number of supercharging stations. We have about 1.55x the number of charging stations compared to Tesla. Wherever our users go, they can easily find our charging stations. We completed the coverage along the G318, Sichuan to Tibet highway route, and core sections of the 9x9 highway network in China. We now operate the largest urban and highway supercharging network of any automaker in China. Second, speed. Our charging network supports 500 Km of driving range with only 10 minutes of charging, just the time of buying a cup of coffee.
It's really, really fast. Experience. Our charging network provides a very integrated experience between the vehicle and the charging stall, with very intelligent charging planning, charging reservation, and plug-and-charge functionality. Combining that with the lightweight charging poles designed to be easily handled with one hand, it's especially user-friendly for female users. In terms of value for money, compared to the owners of other EV brands, Li Auto vehicle owners enjoy a dedicated preferential electricity charging rate, saving on every charge. In addition to that, our owners can use their loyalty points to offset our charging costs. These aren't just numbers on paper. They translate into real everyday charging experience that our users can provide positive feedback on every day.
Our 2C and 4C charging stalls have been open to vehicle owners of other brands, but those other owners won't be able to have the same level of experience as our vehicle users because the best experience requires seamless integration of the vehicle as well as the charging stations. Thank you.
[Foreign language].
In fact, the beginning of 2025 has really been the most challenging period for assisted driving. With technological and experience progress slowing down, at the same time, regulatory pressures have grown. However, I see this as the darkness just before dawn. The VLA represents a clear pathway towards L3, L4, and even L5 autonomous driving. As it operates really in a way like human, only we can expect it to perform 10x or even 100x better than human drivers in the future. VLA is the true agent in the third stage of AI, and we build its capabilities in two ways. First, it acquires human experience by learning from human data and the foundational model through large-scale operation of SFT and RLHF to achieve performance on par with human drivers. It's like an apprentice learning from a master.
More importantly, it undergoes reinforcement training within environments generated by a world model using RLAIF to train agents. For example, we many times see L4 vehicles in China as well as overseas falling to traps, for example, construction nodes. This kind of data is very limited in the real world and cannot be used for training. However, with a world model, we can generate countless variations of cars falling into these traps and enabling the agent to learn rapidly and resolve these issues effortlessly through reinforced practice. The world model provides data that is more difficult, more comprehensive, and higher in quality and more challenging than real-world data. This approach effectively addresses critical issues in real-world data such as underrepresentation, imbalanced distribution, and the impact of unavoidable polluted data on agents.
We believe that ongoing iterative training and development, the VLA will achieve a level of driving safety at least 10x that of a human driver within the next two years or so. The challenge behind this lies in equipping the agent with both a more powerful brain and a stronger heart. By brain, we refer to model scale. The end-to-end model has 0.3 billion parameters, while our VLA model has 4 billion parameters, which remain far inferior to human capabilities. Expanding this scale will substantially enhance generalization capability. Scaling the model will also require significantly more computing power on the device side, which acts as the heart to supply the brain with the necessary resources to supply larger models.
In the meantime, the realm of world models and RLAIF will likely require at least 10x more computing power for reasoning than before to generate the environment and data necessary for RLAIF training. We anticipate that with the enhancements in reinforcement training, model scale, and computing power, the speed of our progress will far exceed anything we've seen in our previous approaches. I believe that in the next 35 years, the largest application of agents in the digital world will be programming, coding, while in the physical world, it's going to be autonomous driving. I believe that level 4 autonomous driving will be achieved by 2027. Thank you.
Thank you. As we are reaching the end of our conference call now, I'd like to turn the call back over to the company for closing remarks. Ms. Janet Zhang, please go ahead.
Thank you once again for joining us today. If you have further questions, please feel free to contact Li Auto's Investor Relations team through the contact information provided on our IR website. This concludes this conference call. You may now disconnect your lines. Thank you.