Duolingo (Minutes): Many future opportunities, actively focusing on DAU expansion

The following is a summary of the $Duolingo(DUOL.US) 3Q25 earnings call organized by Dolphin Research. For an interpretation of the financial report, please refer to "Duolingo: Weak Guidance + Uncertainty of Short-term Changes, Green Bird Falls from High Valuation Altar"

I. Key Financial Metrics

II. Management Report

We achieved excellent results in the third quarter, with all metrics performing steadily. This year, we are poised to deliver outstanding performance once again. Currently, over 50 million people use Duolingo daily. We expect user registrations to reach nearly 1.2 billion this year, with a year-over-year growth rate of 33% and an adjusted EBITDA margin of 29%.

Overall, we have not only rapidly expanded our influence but also improved our profitability. However, we still feel that we are just at the beginning of this journey. We firmly believe that artificial intelligence will fundamentally change the education sector, and we have the capability to create more efficient teaching products than ever before. This is why this moment is so exciting for Duolingo.

III. Analyst Q&A

Q: First, can you help us better understand the driving factors behind the growth in daily active users and overall user engagement in the fourth quarter? Additionally, how far are we from returning to normal levels in U.S. marketing? I know you mentioned increased user exposure, but has this started to translate into user growth? Or how should we view this trend in the fourth quarter?

A: We just announced our third-quarter performance data, with a year-over-year growth of 36%, which we are very pleased with. As for how we achieved such results, there are different perspectives. From a positive angle, such as the long-term partnerships mentioned in our shareholder letter (like the collaboration with Luckin), this indeed helped us significantly increase user numbers in Asia. Additionally, we have made several product improvements that have also contributed to enhancing user retention. Overall, these achievements are very significant.

On the other hand, as we mentioned last time, there was a period when we stopped posting those "exaggerated and absurd" content on social media. We did this because we were listening to user needs and striving to cultivate user affection for the brand. By no longer posting those absurd contents, our posts naturally became less likely to go viral. This change indeed impacted the growth of daily active users.

The good news is, in the past few weeks, we have started posting those "crazy" contents on our social media accounts again. Although the popularity of these contents has not yet returned to previous peak levels, there is indeed a noticeable upward trend. We expect this change to have a positive impact on daily active users.

Regarding daily active users, we expect the growth rate in the fourth quarter to slow down compared to the third quarter, but we are quite satisfied with the current stable situation. It should be noted that we will not provide specific forecasts for daily active users in the fourth quarter; however, it is certain that the year-over-year growth rate of daily active users in September and October reached about 30%. This growth rate was achieved on the basis of strong growth in the last quarter of last year.

Q: You previously mentioned three core areas: monetization, user growth, and overall improvement in teaching effectiveness. Can you explain in detail what prompted us to shift our investment focus towards long-term development? Is it the opportunities brought by artificial intelligence, or should we consider this change's impact on booking growth from other perspectives? I know you stated a few years ago that your goal was to allocate more than 25% of investments to related areas. So, I'm curious, how do we gradually achieve this goal?

A: As I mentioned in the shareholder letter, there are indeed tremendous opportunities at present. We believe that in the coming years, the education sector and the way people learn will undergo fundamental changes, all thanks to the development of AI.

Because of AI, we now have the ability to develop an application whose teaching effectiveness far surpasses any teaching tools humans have ever had. Its teaching level can rival that of human teachers and is more engaging.

Currently, our app has 135 million active users per month. If we can develop such an application with excellent teaching effectiveness, then our user numbers will reach billions, which is our goal.

This is why we are making long-term investments.Specifically, our investments are more focused on improving teaching quality.If our teaching level improves, it will indeed help increase user numbers, but this effect takes time to manifest.

As long as we continue to optimize our products and services, user numbers will naturally increase over time, but this process requires time. Additionally,there is a lag between user number growth and achieving profitability—because people usually take some time before choosing to subscribe to our services. So, this is indeed a process that requires long-term persistence. However, we are very optimistic about this prospect. Our goal is simple:due to the enormous market opportunity, our goal is to maintain rapid growth in daily active users for a long time.

Q: Considering the enormous potential of the opportunities you mentioned, does this shift in long-term priorities indicate a strategic and lasting change, and is this change expected to continue? Or do you plan to return to the existing priority arrangement at some point? Additionally, associated with this new strategic focus, can we expect the booking pressure seen in the fourth quarter to persist into next year, and will this pressure continue?

A: It will take several years before we have such an application, so during this time, we still need to continue investing resources. However, I must say, we have not set specific goals, such as "must complete this project by next year," but we are very excited about many new features that will be launched in the product next year, such as:

(1) We will provide higher-quality video call services for beginners. We call it "guided video calls." The social features of this application will also be significantly enhanced.

(2) Currently, our math course content is still very limited; but in the coming months, we will supplement all K-12 education standard content for math courses.

Our chess course adopts a player-versus-player mode, and it is currently performing very well, with rapid development.We will also comprehensively reform the music course. For the first nine language courses we offer, we will be able to help learners start from scratch and reach a Duolingo score of 130—at this score, learners can find jobs using that language.

For a long time, we have discussed with you and our investorshow to maximize the platform's user lifecycle value.Although the changes described by Luis are currently just minor adjustments, they will help us attract more users in the long term, encourage users to complete more learning tasks, spend more time on the application, and make Duolingo users more active and engaged in the learning process.

Q: At the Duocon conference, you mentioned that the chess course has millions of active users daily. In this regard, have you found any differences in learning enthusiasm or user retention between users participating in the chess course and those using your core language learning services? Additionally, I'm curious about your newly launched PVP feature. Can you tell us when this feature is expected to be fully launched?

A:Chess is currently one of our fastest-growing courses; its growth rate far exceeds that of math and music courses, even surpassing the traditional growth rate of language learning. In fact, our chess users have reached millions. Although we have not disclosed specific user numbers, it is certain that this number has exceeded the number of users in math and music courses.

Maintaining the stickiness of chess game users is not easy—after all, this type of game has not been around for long. Our chess course has only been running for three to four months. However,from the data of the past three to four months, user stickiness is indeed higher than that of language learning applications, so we are very satisfied with this. Additionally, as we mentioned earlier, we have started launching the PVP mode. PVP mode is a player-versus-player mode, which means users can interact with other players in the game.

Currently, 50% of users using iPhone and iOS systems can already use this feature. This feature has not yet been launched on the Android system, but it will soon be available on Android. Therefore, we expect that in the coming weeks, all users who have installed the Duolingo application will be able to engage in PVP chess battles.In the coming year, we expect a significant increase in the number of users related to chess features.

Q: How do you prioritize user growth over profitability? Additionally, what specific adjustments need to be made to the application itself, and from which aspects should I consider these adjustments?

A: When we conduct experiments, we always need to make trade-offs between different factors. Some experiments can indeed improve the performance of all metrics, which is certainly good—because such experiments are clearly worth adopting.

But sometimes, certain experiments may improve one metric while harming another. Let me give a fictional example: Suppose currently, free users receive 25 energy units at the start of each day, and each exercise they do consumes 1 energy unit. If we conduct an experiment to reduce the energy units free users receive each day from 25 to 24, meaning they will have 1 less energy unit available each day.

Although we know that doing so can make us more money because more users will run out of their "energy points" and be forced to subscribe to our services, we also understand that this approach will reduce the number of daily active users because it will frustrate some users.

For such issues, we always need to make various difficult decisions. Butin this decision-making process, we prioritize user growth over profitability.

Q: What changes have you observed that prompted the company to make this shift? In other words, as AI continues to develop, the company is now more focused on long-term development rather than short-term profitability. What exactly led to this shift? Additionally, can you roughly predict how long this shift will last? Will it continue until 2026, or will it be shorter? Can you provide more information?

A: Over the past few years, we have become increasingly aware that we are at a special moment—especially in the education sector: how global education models are changing and the high-quality teaching content we can provide on the Duolingo platform. This can be seen from our own data metrics:through features like video calls, the speed at which we release new content is accelerating, and these teaching effects are significantly improving every month.

This shift has been brewing for some time. In the past one or two months, I have indeed worked hard to get the entire company to recognize and support this shift, and the company has realized that: this opportunity is enormous for us; we should prioritize achieving long-term sustainable development while developing an educational application that is better than any previous product.

As for how long it will take to see results, this is indeed an interesting question. I mean, you might be wondering: will this affect our subscription business performance? And will this impact last forever? I don't think so. It's just that we need some time to see the actual effects of these long-term investments—in other words, the financial returns.

As we have seen in the annual guide and the fourth-quarter report, this adjustment has indeed had some impact on the company's financial situation, but the impact is not significant. So, can we expect this impact to continue until 2026? It's certainly possible.However, I think overall, the financial impact of this adjustment is actually quite limited. And we believe that such an adjustment is worthwhile—because as Luis said, this is indeed a huge opportunity. Therefore, from a risk and reward perspective, such a decision is reasonable.

Q: Regarding the practical significance of these changes, should we slightly reduce our focus on paid conversion to enhance the user experience? Or do we have grander plans, such as incorporating video call features into some lower-priced packages? Additionally, I have some follow-up questions.

A: We will focus more on user growth goals. Perhaps we will try to apply video call features to other levels of packages, but this is not directly related to what we are currently discussing. In fact, we have always been considering how to integrate different features into different packages, and we will test this. During the testing process, we are also working hard to optimize the lifecycle value (LTV) of our platform.

Due to the changes in the past few months (such as the Energy plan), by September, we did observe some temporary changes:the conversion rate from free users to paid users slightly declined, but overall, this conversion rate is still growing year by year.

Q: Regarding the follow-up work, I would like to ask: In the previous conference call, you mentioned the metric of the average number of words spoken per call. This is actually a new metric you plan to start testing. So, what progress have you made on this metric so far?

A: This is a very important metric, and we have indeed successfully improved it. Understanding this situation about our company is very important. Whenever we focus on a specific metric, we can always drive it in a positive direction.Regarding this specific metric, the average number of words spoken by each Max subscriber this year has more than doubled, and we are very pleased with this.

Regarding the upcoming video call feature improvements, there is one aspect that excites me very much: the current video call feature only supports communication in the language the learner is learning. For example, if you are learning Spanish, then all content in the video call can only be conducted in Spanish. While this is certainly helpful for practicing Spanish, it is too difficult for beginners. If you only know 20 Spanish words, then having a complete conversation in Spanish is almost impossible.

Therefore, what we are testing now is a feature called "guided video calls," which basically supports bilingual communication. In other words, if you are an English speaker learning Spanish, then in such a call, part of the content will be expressed in English, and the other part will be expressed in Spanish. This approach is much easier for beginners. We found that when beginners use this feature, they actually say more words in each call because they can really participate in the actual communication.

Therefore, we believe this will indeed help drive users to switch from other platforms to using Max. Additionally, I have two points to clarify:

First, most of our users, especially those who subscribe to Max, are beginners. Therefore, we firmly believe this will greatly promote large-scale user conversion; secondly, regarding these new features, we have not yet started promoting them, nor have we used them to attract users to subscribe to Max. We have just made these technologies public, and the next step is to let those who have not yet subscribed to Max know about these technologies, thereby encouraging them to choose to subscribe. We are very excited about the positive impact these technologies can bring to Max.

Q: From the perspective of AI, it indeed makes education more efficient, at least from what you described. So, has this accelerated your progress in developing new language learning modules? I know you already have some related modules, but will you be involved in new learning areas next? What changes in the field of artificial intelligence have enabled you to do these things now?

A: These situations have indeed prompted us to accelerate the advancement of language teaching projects. However, this does not mean we will add more language courses. In fact, people have clear preferences for which languages they want to learn.

Currently, we offer teaching courses in 40 languages,and the demand for languages not included in the teaching range is very low. However, the nine languages we teach most frequently—such as Spanish, English, French, German, and Italian—are the languages people most want to learn. These nine languages account for the learning needs of the vast majority of our users.

Therefore, you will see that we will accelerate the pace of adding new content for these nine major languages. However, for most of these languages, we have not yet helped users reach our goal—that is, to have users achieve a Duolingo score of 130, which is equivalent to the B2 level of the CFR language level, and only at this level can users find jobs with their language knowledge.

In the coming months, you will see us adding corresponding learning content for all nine major languages.Additionally, you will find that we have adopted more different modules in language teaching methods, and these modules will be more intelligent in the teaching process—that is, they can better adjust according to individual learning situations. At the same time, we will use artificial intelligence technology more in the teaching process to achieve more free response sessions, allowing teaching content to better meet the needs of each learner.

Regarding other subjects, we will also use AI. In the field of mathematics, we are using this technology extensively to create more teaching content. As for whether it will expand to other subjects, we have not yet started working on this aspect. However, I do not rule out the possibility of adding other subjects next year. Just like when we developed the chess teaching system, it took a full nine months from conception to actual launch. Therefore, although the possibility is not high, we do have the possibility of adding content for other subjects next year. However, we are not currently researching projects related to other subjects.

Q: Regarding AI, can you talk about your observations on computing power costs and gross margins? Is the actual situation better than expected? Is the speed of this change faster than you imagined? What impact will this change have on the industry? One important question we are often asked is: what is the trend of AI development in the current competitive landscape? What are your thoughts on this?

A: Regarding cost issues, costs have indeed been declining. But for us, optimizing costs is not the top priority. Currently, our primary goal is to provide users with the best possible experience—because it is these users who teach us the most and give us the strongest sense of engagement.

Most of our AI features—at least the most costly ones—are concentrated in the "Duolingo Max" product; and the price of this product is high enough, so for us, using these AI technologies is still profitable. Because of this, we will not deliberately further optimize the costs of these features. Currently, Max accounts for 9% of our total user base; in the third quarter, bookings from Max doubled year-over-year—from this perspective, Max's performance is quite good. However, its performance still has not reached our original high expectations.

Although AI costs have decreased, I do not want to give the impression that we are no longer willing to invest. As Luis said, we are now at an excellent time to seize huge development opportunities.We expect the adjusted EBITDA margin this year to reach 29%, which is very close to our long-term adjusted EBITDA margin target.Therefore, in achieving this goal, we will not be afraid to invest in innovation. We will gradually increase our investment in these areas.

Regarding the competitive landscape, people often comment: Regarding the competitive landscape in the field of artificial intelligence, people mainly mention two points: first, since you can use ChatGPT to learn languages, why choose to use Duolingo?

Actually, we are not too worried about this issue. We have said before, our advantage lies not only in teaching quality but also in continuously attracting users to maintain their interest in learning.

To learn a language, you need to invest years of time. Truly mastering a language requires daily persistence, and we must ensure that you can continuously engage in the learning process. Not only that,we also need to develop a multi-year learning plan for learners.However, with ChatGPT, you can ask it for some words, but it cannot provide you with a complete, multi-year learning course. So we are not too worried about that issue.

Additionally, some people worry that because of real-time language translation technology, no one will be willing to learn a new language anymore.

However, we are not too worried about this either. I believe that at every Google I/O conference in the past 10 years, they have demonstrated real-time language translation technology. They do this every year, and this technology is indeed very practical and effective.However, despite the existence of such technology for 10 years, we have not seen any decline in people's interest in learning foreign languages.In fact, people's interest in learning foreign languages has increased.

The biggest reason for this situation is: Our users can be roughly divided into two categories.One category is people who treat language learning as a hobby.For these users, whether our computers have the relevant functions is actually not important—by the way, their situation in learning languages is the same as in learning chess. Although computers are far superior to humans in playing chess, there are still millions of people willing to learn chess. Therefore, for those who treat language learning as a hobby, whether computers have the relevant functions is actually not important.

The other major category of people learning languages with us are those who truly want to learn English.For them, using a phone that needs to be held up to learn English is not the way they want. Therefore, we are not too worried about this. It's just that some people like to talk about these things on Twitter... but we are not bothered by it.

Q: I know you did make some price adjustments or increases earlier this year. So, what kind of feedback have you received regarding these adjustments? Secondly, I remember you originally planned not to introduce more price adjustment measures temporarily—mainly for the standard packages launched for new users. So, what are your plans for future price increases? Considering that you will focus on user growth in the fourth quarter and beyond, I think you probably won't raise prices now. However, I would like to hear your thoughts.

A: Regarding pricing, we will conduct corresponding tests—we will try various pricing strategies. Ultimately, we will launch products that we believe are beneficial to the entire platform. Although it is not yet possible to determine which plans will ultimately be adopted, you will definitely see us testing prices. The "test prices" here refer to attempts to raise or lower prices.

We may try to launch some products with significantly reduced prices or launch some discount packages. What I am saying here are just some measures we might take, but if the test results are not ideal, we may not actually implement these plans. However, we are likely to launch some lightweight products priced at only half the original price.

We have often discussed improving sales performance by raising product prices, and we will indeed do so. Luis just mentioned that we will continue to try this method, and this approach will continue.This year, our average revenue per user (ARPU) has been steadily increasing each quarter, with a growth rate of about single digits.This change is also reflected in our sales forecasts, and this growth is mainly due to Max's efforts, rather than product price adjustments.

In the past, we have indeed adjusted product prices multiple times, and these price changes do affect the discount range in our only promotional event of the year. Therefore, if the product price is high, we need to try different discount strategies.Every fourth quarter, when we discuss promotional plans,we mention this point—because the fourth quarter is the most unstable sales quarter, after all, we will launch New Year promotional activities in this quarter.

This year, we adopted the "Energy plan" as the core pricing mechanism to carry out the New Year promotional activities. Therefore, before the end of this year, we will try various methods to design this promotional activity: including how to carry out the activity, how to display information, and the best timing for the promotion, etc. All these contents are also detailed in this guide.

Q: I want to confirm that the main reason for the slowdown in order volume and revenue growth in the fourth quarter is that the company temporarily slowed down the conversion rate of users from free accounts to paid accounts, instead focusing on user growth? I just want to ensure that this is indeed the main reason for the slowdown in revenue growth, rather than a decline in the retention rate of paid users or an increase in the churn rate of paid users due to price increases.

A: That's right, these changes in the fourth quarter are mainly because we started to shift our focus to some long-term projects.This means we need to focus on user growth while also investing more resources in certain areas—we not only prioritize user growth but also invest more effort in improving the teaching quality of the product; for this, we had to divert some funds from resources originally used for monetization.

Q: When launching these product updates, do you specifically design for certain groups of people (such as Gen Z)? Gen Z's learning methods seem to differ from millennials. So, how should we evaluate whether these product updates are really effective? Specifically, what metrics will you use to prove that these updates have indeed achieved the expected results? User growth can indeed bring benefits—some changes do help promote user growth, but this does not always mean that those "sticky users" will convert into truly active users. What are your goals for these changes? How will you measure whether these changes have been successful?

A: Indeed. Different people have different learning methods. But you will be surprised to find that the similarities between people are actually more than the differences. In fact, this difference exists not only between different generations but also between different regions.

We repeatedly find that not only do many people's learning methods are roughly the same, but the factors that encourage people to use the product more frequently also show a high degree of similarity in different regions. In other words, this effect exists in every country—so there are many commonalities between these factors. Currently, our main focus will be on creating better learning courses for the public.

These courses will indeed be adjusted according to the specific situation of each student. One of the aspects that need the most adjustment is the difference in learning speed. In fact, as we age, learning speed does slow down. I deeply understand this point—as someone who is getting older and whose learning speed is gradually declining, I can say with certainty that this situation is common and not controversial.

So, regarding the issue of work pace, it is actually easy to adjust. We can indeed easily adapt to different work paces, and we have been doing this for quite some time. As for the metrics we need to focus on, user growth is undoubtedly a very important metric—we are currently focusing on this metric, and it is the most important metric in the company.

Regarding improvements in teaching methods, as I mentioned before, these improvements will not immediately lead to an increase in user numbers, but they require time for users to experience and spread by word of mouth. Since we started conducting annual evaluations, Duolingo's teaching effectiveness has indeed been improving year by year. We are likely to see further improvements in our teaching methods, and the speed of these improvements will be faster than in the past because we are now investing more effort in improving teaching quality. Therefore, our assumption is:these improvements will eventually translate into an increase in user numbers. However, this growth process may not be linear, nor will it happen quickly in a short time.

Q: I know you may not want to talk about business planning for 2026. But obviously, the company needs to invest a lot of funds in various businesses and content development. Can you roughly tell us how large we should expect these investments to be? Additionally, do you still have expectations for the company to continue improving EBITDA margins in the coming years?

A: We will discuss this topic again in February. However, I must add that while expanding the platform scale, we have also made remarkable progress in achieving long-term profitability goals. Currently, our platform has 135 million monthly active users, and we expect the total order volume this year to be close to $1.2 billion.

Therefore, we are working hard to expand our business scale, ensure discipline in the operation process (cost control), and improve profit margins.While doing these things, we are not afraid to invest. We believe that even with investments, we can still maintain high profitability.But from now on, we will no longer make linear expansion of profit margins the primary goal—because we believe the market opportunity is too great, so we should prioritize investment. This is our idea.

Q: Regarding advanced English learners and how to increase their learning enthusiasm and encourage them to continue using these learning platforms. I want to know, has AI accelerated this process? Have the points we initially proposed been verified? Additionally, has the slowdown in learner growth affected these advanced learners? Do you have anything else to say about this group?

A: It was probably a few years ago when we started discussing this topic. Learning English is indeed a huge opportunity. In fact, 80% of people worldwide are learning English. Therefore, we began investing resources to provide higher-quality teaching services for English learners.

This year, we have launched several new English learning projects here. Now, all our English courses cover helping learners reach a Duolingo score of 130, because once they reach this score, learners can use this language to engage in related work.

We have indeed achieved such results: we are continuously making progress in improving people's ability to learn English, and from various data, it can be seen that the number of people learning English, especially those with higher English proficiency, has been steadily increasing. Additionally, I want to say that,the fastest-growing regions are those that vigorously promote English learning. Asia is a good example.

We do continue to release various "exaggerated" content marketing in Asia, whether through social media or other channels; at the same time, we are attracting more and more users there, which is the fastest-growing part. We have said before that this process takes some time—after all, people need some time to truly realize that Duolingo is indeed excellent in advanced English learning. Although I think we have made good progress in this regard, we are still some distance from our goal; in fact, even next year, we will see the number of users learning advanced English continue to increase.

Q: Regarding the Family Plan, can we get some latest progress information? I know you have been adding various new features to this plan and are also committed to promoting its wider adoption. So, what stage are we at now? How far can this plan develop in the future?

A:The "Family Plan" service is still performing well, with about 29% of users using this service in the most recent quarter.Therefore, the number of users of this service has indeed grown significantly. Moreover, I believe there are many reasons to believe that its user numbers will continue to grow steadily.

The reason we were able to achieve such rapid growth last year, with a year-over-year growth rate of 42%, is partly because the family plan promotional activities launched during the New Year period achieved very good results. Therefore, we have reason to believe that we still have a lot of room for development in the family plan business area.

Q: Regarding user feedback on the Energy plan and how they have influenced some of the market trends we discussed this quarter. Additionally, when we consider the entire product development process, how much of it truly depends on the quality of the product itself, and how much is related to the cultural influence Duolingo had earlier this year? Also, why did the product development speed suddenly slow down after encountering social opposition?

A: The performance of the Energy plan fully met our expectations: it not only increased bookings but also improved daily active users. This is indeed a very good result. Our promotion strategy is: first launch these features for iOS users, and then for Android users.

Over the course of a few months, we gradually rolled out this feature to every relevant platform. Now, this work is complete. I don't know exactly how many weeks ago it was, but now, all users who have updated the application in the past six to nine months can already use this new feature. We are very satisfied because this feature has indeed achieved our expected results, and it is now officially launched.

Regarding the balance between product and marketing you mentioned, both are crucial for us to achieve business growth. In the long run, if we can develop products with high user retention and effectively convey product value, that is undoubtedly our ideal goal. Therefore, we are currently and will continue to focus most of our efforts on product development and marketing work.

Cultural influence is indeed important. But for me, culture is just one factor driving things forward; what truly determines success or failure is the product itself.

In response to social influence, we have increased our marketing investment in the U.S. region, although the amount involved is relatively small, but it is currently yielding good results. We will continue to implement this strategy in the fourth quarter. As we discussed earlier, the economic growth rate in the U.S. is slower than in other parts of the world, partly because the economic growth rate in other parts of the world is too fast.

Q: The first question is about the market situation in China. Can you talk about the trend of engagement in the Chinese market over the past few quarters and throughout the year? What are your expectations for this? Additionally, can you comment on whether Max might enter the Chinese market this year or next year? The second question is about Max's renewal rate. Please talk about the trend of renewal rates since the third quarter and what your expectations are.

A: Our business in China is developing very well. I think China is our fastest-growing market;in terms of daily active users, China is also our second-largest market and is still growing rapidly.Currently,the revenue share of the Chinese market is still relatively low, accounting for only about 5% to 6% of our total business,but the growth momentum is very strong, and user engagement and retention rates are also very high.

Max is currently being tested in China. Because this is a large language model, we need permission to test in China.The testing work is ongoing,and this product is expected to be officially launched in a few weeks or months.However, the specific time is difficult to determine because it does depend on the results of the relevant approval process, but the overall progress is relatively smooth.

Regarding China, I want to say that this is indeed a very significant opportunity for us. Of course, China also comes with certain risks—such as geopolitical factors. Therefore, our response strategy is: we have not invested huge amounts of money in marketing activities in the Chinese market. Our investment in this area is very limited, but even so, our business has still achieved considerable development.

So, this is indeed a very good opportunity. But if something unexpected happens at some point, then we may ultimately not invest so much money there. That's what we plan to do.

Regarding MAX's renewal performance: Although Max's renewal service is very important, from a broader perspective—I remember someone asked about user retention across the entire platform before—this situation remains stable and has not undergone substantial changes. As for Max itself, we mentioned in the last conference call that from the third quarter to the fourth quarter, more users will complete the Max renewal process.

Additionally, what we observed is:currently, Max's performance is still relatively weak because its performance improvement speed is relatively slow; however, its performance is indeed slightly better than Super.However, it is still too early, and the fourth quarter is the real test of its performance. Therefore, we will discuss this issue again in the next call.

Q: From a geographical perspective, which countries or regions do you think have the most growth potential in the next quarter? Another question is related to the Duolingo performance data you released this year. Because this data actually reflects learning outcomes. So I want to know, what is the relationship between this way of measuring learning outcomes and core user performance indicators? Additionally, will this measurement method become a key driver for attracting more users in the next quarter?

A: From a regional perspective, the fastest-growing region for us is Asia, and we expect this situation to continue for some time. Although I cannot predict the future, I believe this trend will continue in the short term. Of course, China is leading in this regard, but it is not just China; in fact, the entire Asian region is growing at a very fast pace.

Regarding scores, we are very excited about this. What we announced is: first, all our mainstream language courses now support displaying Duolingo learning scores; in addition, users can also share their Duolingo scores on LinkedIn. We found that many people do share their scores on LinkedIn. This shows that they do use this score for practical purposes such as job hunting, and we have indeed seen such application scenarios.

The score range of this test is also comparable to our Duolingo English test results, at least for English learners. And English learners are the ones who truly care about this type of test score, especially English learners in Asia. Therefore, we are very excited about this result.

Its ultimate goal is to become the learning standard for at least major languages, especially for English. This way, when people ask you "how much French do you know" or "how much English do you know," they will no longer get vague answers but clear, standard-based evaluation results.

For example, some people might say: "Oh, my French level is intermediate." But we hope people can express it like this: "My French level has reached 60 points on Duolingo, or my English level has reached 80 points on Duolingo." This is what we really want to see. And we believe we have already made considerable progress in this regard.

Regarding language scores, especially English scores, the Duolingo English test is now recognized by more than 6,000 educational institutions worldwide. These institutions include all Ivy League schools and 99 of the top 100 universities in the United States. We believe that with this reputation advantage, coupled with the large user base we have in the app and initiatives such as content sharing through platforms like LinkedIn, we are expected to become the industry standard for measuring user skill levels.

<End Here>

Risk Disclosure and Statement of this Article:Dolphin Research Disclaimer and General Disclosure