At the beginning of 2026, Alibaba announced a brand upgrade for Qianwen and concurrently revealed product plans for Qianwen AI glasses, a smart ring, and AI earbuds. On March 2nd local time in Spain, at MWC 2026 in Barcelona, the Qianwen AI glasses made their global debut. In addition to domestic media, this was the first exposure for many international media outlets to Qianwen's glasses: numerous foreign journalists were seen queuing at the experience zone to try them on, testing features like translation, photo recognition, and task completion via voice commands, with the area being fully occupied throughout the morning. On the same day, the Qianwen AI Glasses G1 began pre-orders in China, with the first batch of "one-command task" capabilities scheduled to be activated via an OTA update by the end of the month. The pace is exceptionally rapid.
Choosing MWC as the global stage for the debut sends a clear signal: Qianwen AI glasses are not solely targeting the domestic market; official sources indicate plans for a full-scale entry into the international market within the year. If the Chinese New Year period represented Qianwen's battle for app entry points, then MWC may mark the starting point for its campaign for terminal entry points. During MWC, Song Gang, head of AI hardware for Qianwen's consumer business group, engaged in an in-depth discussion. This hardware veteran with nearly two decades of experience in the mobile phone industry is attempting to answer an ambitious question: What changes will occur as Alibaba begins to heavily invest in AI hardware?
The conversation first turned to branding. Late last year, Alibaba internally decided to consolidate all AI hardware under the "Qianwen" brand. When asked about the rationale for this adjustment, Song Gang's response was succinct: "The assistant adjusted first, then the hardware followed." This might sound simplistic, but upon closer consideration, it highlights a sequence: it's not about developing hardware and then finding a brand to attach to it; rather, it's about the assistant's capabilities expanding outward, with the hardware evolving to support that expansion. This sequence defines the fundamental difference between Qianwen glasses and most smart glasses on the market. The logic for most other hardware brands defining smart glasses is hardware-defined scenarios, whereas for Qianwen, it is AI content and services driving the hardware.
"Hardware development has reached a relative bottleneck; further progress will be more driven by AI," Song Gang stated calmly, but his meaning was clear: while Qianwen glasses maintain leading hardware design and configuration, the real incremental value lies not just in the temple design or display, but in each model iteration of the Qianwen AI assistant and every newly integrated ecosystem service. The viral success of the "one-command food delivery" feature on the Qianwen app during the Chinese New Year period somewhat validates this logic—once the app-side capability is proven, the glasses side must quickly incorporate it. Song Gang said this would be implemented via OTA by the end of March.
The essence of the brand upgrade is to redefine a hardware product as a physical carrier for AI services. Qianwen's ambitions clearly extend beyond just glasses. Concurrently with the MWC launch, plans for the ring and earbuds were also revealed. Why these categories? Why not a smartwatch? Why not a more mainstream form factor? Song Gang provided three keywords, which he sees as three layers of supplementation that AI hardware provides relative to smartphones:
First, and most importantly, always online. A phone in your pocket—if a notification pops up while you're driving, you often can't see it in real-time. Glasses on your face and earbuds in your ears are naturally closer to the user. Second, proactive services. "In the past, when we talked about smart features, it was often human-made intelligence—setting an alarm to remind yourself to perform an operation. In the future, it should be reversed: the machine reminds itself, then prompts you to make a choice or decision." He gave an example: you're running a marathon; after two minutes, the ring can assess based on biometrics whether you can finish today, and if your condition is poor, it will alert you early. Third, more sustained collaboration compared to a phone. You take off your glasses at night, but the ring continues monitoring heart rate, blood oxygen, and sleep, feeding data back to the AI assistant. When you put the glasses back on the next morning, the assistant already understands your sleep quality better than you do and can provide a report and reminders. Regarding the ring and earbud categories Qianwen is also developing, the ring is not just for health monitoring but also acts as a "remote control" for the glasses—introverted users who don't want to speak to wake the voice assistant can press the ring to ask a question directly, or double-click to take a photo. The earbuds target users who don't wear glasses. As for why not start with a smartwatch? "A watch still requires you to lift your wrist and swipe," Song Gang gestured, "the ring just requires a press; it's less intrusive for the user." All category choices point to the same judgment: prioritize the form factors that can extend the capability boundaries of the AI assistant.
However, among all categories, glasses have the highest priority. "Glasses can capture 80% of human information; their significance for understanding and assisting people is the most comprehensive." The others are extensions of the glasses.
Discussing market performance, feedback for the current Quark glasses has far exceeded expectations. Simultaneously, Song Gang revealed another data point: the number of AI interactions per glasses user is six times that of third-party AI assistant apps. "We measure AI hardware more from an interaction perspective. Each wake-up represents helping the user accomplish a task. A six-fold increase is very positive." Using an AI assistant on a phone requires taking out the phone, unlocking it, opening the app, and inputting a command—this series of actions creates an invisible barrier; often, you think of asking something but then decide against it. But with glasses on your face, you just speak. The collapse in interaction cost directly leads to an explosion in usage frequency.
An interesting detail: during the Chinese New Year, a media professional drove home wearing the glasses the entire way, asking questions about anything they saw. Upon arrival, they counted over 200 interactions. "When we get accustomed to glasses, we realize that taking out a phone is a high-cost action." But the significance of the six-fold interaction increase isn't just "more usage." The deeper logic is: the more interactions, the deeper the model's understanding of the user. The glasses see what you see, hear what you hear, and know if you're running or on a high-speed train. Hardware is not just an output terminal for AI capabilities but also an input terminal for multimodal data. "There's a world of difference between an assistant that understands you and one that doesn't," Song Gang said.
A point of potential misunderstanding needs clarification: Would this data be used to train the model? Song Gang's response was clear: "This is the assistant's contextual memory of the user, which is separate from model training." The data makes the assistant understand you better at the interaction level but is not extracted to optimize model parameters.
The buzz around the "one-command food delivery" feature during the New Year period showed the entire industry how new interaction methods can help software ecosystems efficiently reach entirely new user segments. Over 4 million users aged 60 and above placed their first food delivery orders using voice—they had never used instant delivery services before. This isn't a migration of existing users but a new interaction method unlocking a completely new demographic. The impact on the hardware team was direct. "We previously thought we could afford to be less hurried, but seeing such high demand now, we must move quickly." By the end of March, the "one-command task" feature will be activated on the glasses via OTA.
Song Gang offered a pointed assessment of this feature, perhaps the sharpest statement in the entire conversation: "One-command task completion is even more of a necessity on glasses." Why? GUI Agents on phones essentially still require you to pick up the phone, unlock it, open an app, and speak a command—the process isn't fundamentally shortened. But on glasses, the flow is wake-up, speak, complete—the chain is compressed to its shortest, making voice the most natural interaction method. "One-command task completion is genuinely a scenario better suited for glasses."
Integration with the Alibaba ecosystem is accelerating. The Amap team is already brainstorming scenarios: in the future, a user wearing Qianwen glasses looking at a restaurant could say, "Qianwen, help me get in line and take a number," which might be more convenient than pulling out a phone. Taobao, Alipay, Fliggy—these Alibaba services are being natively integrated into the glasses via MCP (Model Context Protocol), not through indirect Agent calls, but through direct native service connections. This is a core differentiator for Qianwen glasses compared to many competitors. Many AI glasses emphasize photography, display, or lightweight AR; Qianwen emphasizes "life cycle closure"—seamlessly integrating with ecosystems like Taobao, Amap, Alipay, and Fliggy. "In the past, if voice was slightly better or photography slightly better, users wouldn't notice without comparing two products. But starting this year, task-handling capability is written into the Qianwen glasses specifications—something others can't claim."
On the Quark glasses, we've experienced how Alibaba has solved many design challenges that previously troubled the industry—dual optical engine display, dual-chip dual-system architecture, hot-swappable batteries, titanium alloy integrated injection molding for the temples combining soft and hard angles, Super RAW low-light algorithm—these aren't developed from scratch but represent the transfer of over a decade of mobile phone hardware expertise. "Glasses are even more complex than phones; those without deep accumulation certainly can't do it well," Song Gang noted.
In terms of design philosophy, Qianwen and Meta have taken completely different paths. Meta's classic model is a sunglasses style, leaning more towards outdoor use. Qianwen aimed from day one for "all-weather, everyday wear." This decision drives subsequent requirements: they must be aesthetically pleasing, comfortable, and have sufficient battery life—hence the hot-swappable battery design. "Glasses aren't suitable for taking off to charge. You can't stand being without them even for a minute." The G1 model showcased at MWC continues this approach—8mm ultra-thin temples, approximately 40g weight, along with sunglass versions and 7 colorful lens tint options, with frame styles increasingly resembling traditional eyewear. "We might be the most aggressive in the number of frame styles launched at once among smart glasses. Meta will gradually add more sizes and styles, but the number of frame shapes we launched initially is comparable," Song Gang commented regarding the competitive landscape for smart glasses.
Offline channels are also expanding rapidly. Deep collaborations are in place with major optical chains like Bolon and Baodao, expanding to over 1,000 retail locations. Combined with Alibaba's own e-commerce platforms—channel distribution is indeed an inherent advantage for Alibaba.
Technically, there is a clear division of labor: low-latency and privacy-related processing is handled on the device side, such as image pre-processing and front-end voice recognition; multi-intent understanding, precise Q&A, and third-party service calls are handled in the cloud. "The direction is device-cloud collaboration." In terms of ecosystem openness, NetEase Cloud Music and QQ Music are already integrated, and a tri-end SDK (glasses, phone, cloud) is under development. The approach overseas will be more open—comprehensive output of software and hardware capabilities, allowing developers to customize for their specific scenarios.
Song Gang also revealed a noteworthy signal: Qianwen developing its own hardware is, in a sense, setting an example for the industry. "The near-eye navigation system for glasses was defined by us; the industry only had it after we defined it. We are developing it comprehensively and contributing it to the industry, aiming to empower traditional eyewear manufacturers in the future, enabling them to integrate smart capabilities as well." The goal is not necessarily to profit directly from hardware sales but to understand the needs of the hardware industry through building hardware, thereby optimizing the model's ability to serve hardware scenarios. This reflects Alibaba's platform DNA projected into the AI hardware domain.
Expansion overseas is an unavoidable topic. Choosing MWC for the global debut is itself a signal. The Qianwen brand is launched globally, with glasses planning a full-scale entry into the international market within the year. However, the logic overseas is completely different from China. Domestic users buying smart glasses are AI-driven—they are specifically seeking the "smart" capabilities. Overseas markets resemble traditional eyewear consumption more closely; Meta's sales of over 7 million units rely heavily on the Ray-Ban brand and offline channels. Qianwen's strategy is a step-by-step approach: first, leverage hardware design advantages and the reputation of the Qianwen model among overseas developers to gain a foothold, then gradually build localized service capabilities. Feedback at MWC validates this path—many foreign journalists came to see the glasses due to the知名度 of the Qianwen model.
Regarding competition, Song Gang does not believe one player will dominate. "Overseas, it appears to be Meta, but Google, Apple are investing significantly, and Samsung is also following. There are even more players domestically—car manufacturers focus on car control, Xiaomi on controlling cars and homes, Alibaba on task handling. Each defines its focus differently." He compares it to the early days of smartphones: "It should be a state where a hundred flowers bloom."
Finally, Song Gang, on behalf of Qianwen, summarized the expectation for the smart glasses market in 2026: the year of explosion. Overseas, Meta shipped over 7 million units last year, with predictions of 18 million this year; domestically, the market was around 500,000 units last year, expected to surpass 2 million this year. The Ministry of Commerce has introduced national subsidy policies for smart glasses, and major manufacturers are collectively increasing investment. "The inflection point will arrive earlier, and it's already here," Song Gang stated.
The next wave of users Song Gang看好 are business professionals—meeting minutes, real-time teleprompter, instant translation—these are essential scenarios that can directly replace existing tools. "The early majority represents a market potentially ten or a hundred times larger than the geek player segment." Regarding shipment numbers for 2026, while Song Gang did not mention specific figures, he indicated that with the gradual maturation of the supply chain, advances in color waveguide optical engine technology, and control over core resources, shipments would see "a very significant increase compared to 2025." "Qianwen hardware is a strategic project for Alibaba, and the industry chain will also have more confidence to commit better resources to cooperation," he added.
Towards the end of the interview, Song Gang made a statement that could be the key to understanding Qianwen's entire AI hardware strategy: "AI hardware shouldn't be defined by traditional hardware thinking. It should be the reverse—what scenarios do users need, how can AI assist people in those scenarios, and then, based on that, define what the hardware should be."
On one side is the Qianwen large language model, on another is the task-handling ecosystem of the Qianwen app, and on the third is Qianwen AI hardware—Alibaba's AI advancements on these three fronts converged preliminarily in early 2026 in Barcelona. When model capabilities, hardware design prowess, and life service ecosystems are woven together, this device on the bridge of the nose may truly be becoming the closest AI entry point for everyone, succeeding the smartphone.