The Evolutionary Leap for AI Smartphones

Amidst the sunlit splendor of Shanghai, WAIC 2026 is in full swing on its second day.

Spanning three locations and four venues—Expo, Zhangjiang, and West Bund—this year's event has attracted over 1,100 enterprises, all showcasing a treasure trove of new products and technologies. Leikeji has dispatched a seasoned reporting team, led by senior editor @Chongjia, who is providing live, on-the-ground coverage.

(Image Source: Leikeji)

Throughout the event, robots have undeniably stolen the limelight, captivating crowds with their ability to walk, throw punches, and even serve coffee. However, after exploring the venue, our reporting team found themselves more intrigued by several seemingly ordinary smartphones.

Nubia has seamlessly integrated the Doubao mobile assistant into its latest NaviX Ultra, StepFun has unveiled its STEPX Neo at WAIC for the first time, and Honor is gearing up to let its Robot Phone autonomously turn its head, track people, and follow shots. Outside the venue, OPPO's Xiao Bu assistant has already forged a connection with Alipay's Abao, and whispers in the industry suggest that WeChat is advancing similar collaborations with multiple smartphone manufacturers.

As I watched these phones in action on the screen, it dawned on me that manufacturers are no longer content with AI as just a tool.

They envision Agents as the stars of the next generation of smartphones.

AI smartphones of the past two years were essentially "smartphones plus AI."

The gallery app handled photo editing, the recorder app managed summaries, and the notes app took care of writing—each function neatly compartmentalized. The problem? These capabilities could easily be replicated by standalone apps like Doubao and Qianwen. A copywriting assistant developed by manufacturers over six months might actually pale in comparison to simply switching to another app.

More critically, these functions didn't fundamentally alter how we interact with our phones. AI could engage in a chat or help draft a travel itinerary, but the grunt work of booking tickets, finding hotels, and filling in information still fell squarely on the user's shoulders.

(Image Source: OPPO)

Manufacturers had toyed with the idea of letting AI handle tasks independently, but at the time, on-device models grappled with task planning and tool invocation stability. Sending user data to the cloud raised security concerns, so these ideas were shelved.

Now, with seven on-device generative AI services—including Apple Intelligence, Huawei Xiaoyi, OPPO AndesGPT, vivo BlueHeart, Xiaomi HyperAI, Samsung Galaxy AI, and Nubia's Doubao large model—completing regulatory filings, on-device models have officially transitioned from conference concepts to compliant deployments.

Post-filing compliance, manufacturers are now focusing on comparing Agent capabilities.

This is starkly evident in Nubia's "world's first Agent smartphone," the Nubia NaviX Ultra (Doubao 2nd Gen), developed in collaboration with Doubao.

Unlike its predecessor, the M153, the NaviX Ultra sports a minimalist design with black, pink, silver, and blue back covers, all adorned with a lowercase "nubia" logo. It features a dedicated orange physical AI shortcut button on the right side. Except for the blue model, the other three versions boast frames and modules that match the body color, significantly enhancing its aesthetic appeal.

(Image Source: Leikeji)

The selling point, as everyone should know, is the Doubao mobile assistant.

Last year's technical preview version of the Nubia M153 showcased how an Agent could understand the screen, simulate clicks, and perform cross-app operations. However, that product felt more like an engineering prototype designed to prove technical feasibility—no pre-launch teasers or events, and it even sparked controversy due to its overly advanced concepts.

This year, Doubao and Nubia are bringing this capability to mass-market products, shifting the focus from "AI can operate the phone" to "can it reliably complete tasks for users over the long term."

As Nubia President Ni Fei explained, rather than piling on more AI features, this new Agent smartphone aims to proactively complete tasks based on user needs, reducing the need for users to switch between functions and enabling the phone to handle more operations that would otherwise require manual input.

StepFun's STEPX Neo follows the same philosophy.

(Image Source: StepFun)

Unlike ordinary smartphones, this product utilizes the personal agent Amoo as its primary interaction interface. Whether it's finding photos, setting schedules, planning trips through system apps, or calling third-party services like Meituan, Ctrip, or Alipay, everything can be handled by the same Agent.

Under Amoo's coordination, users don't need to first determine whether a request falls under office, travel, or lifestyle services—they just state their goal, and Amoo decides whom to contact.

In my opinion, this is one reason why smartphone manufacturers are now emphasizing on-device models.

Since Agents need to reside in the system long-term, constantly interpreting screen content, device status, and personal information, they can't rely on sending every action to the cloud and waiting. Keeping simple tasks local and offloading complex ones to the cloud improves response speed while preventing users' photos, schedules, and files from constantly moving between servers.

Of course, being capable and being reliable are two different things.

Successfully completing a dozen steps in a demo might not be too difficult, but whether these Agent smartphones can continue running smoothly amid pop-ups, network fluctuations, or sudden changes in user requests is the real test.

Traditional smartphone assistants have a glaring issue: they won't act unless summoned.

Even if you rush into an airport with luggage, with location data knowing where you are and your flight saved in the calendar, the assistant will likely stay quiet in your pocket, respecting your "fate."

Why? Past smartphone AI relied primarily on text, voice, and screen information—it only acted after you spoke. For AI to provide proactive services, phones need stronger environmental awareness, turning cameras, location, microphones, and sensors into the Agent's eyes and ears.

At WAIC 2026, AI smartphones finally began looking beyond the screen, developing the ability to proactively assess users' scenarios.

Honor's Robot Phone, showcased at the event, is the clearest example.

(Image Source: Leikeji)

In terms of specs, the Honor Robot Phone features the fifth-gen Snapdragon 8 Supreme Edition, a 6.3–6.4-inch 1.5K equal-sided straight screen, 120W fast charging, and comes in Moon Shadow Gray and Star Trail Silver. It boasts a 200MP F1.6 4D gimbal main camera, a 50MP ultra-wide lens, and a 200MP periscope telephoto lens.

The product's selling point needs no introduction—its dual-axis mechanical gimbal, which extends from the body and rotates freely, is unique in the smartphone industry.

Traditionally, smartphone cameras pointed in one direction, with framing and subject selection left to the user. The Robot Phone's lens extends from the body, actively rotating to track subjects and adjusting angles without losing tracking effectiveness.

(Image Source: Leikeji)

This structure isn't just for tracking and framing—it gives the phone the ability to proactively observe its surroundings.

During a video call, the camera follows you as you move; when shooting action footage, it acts like a mini gimbal camera mounted on the phone, handling tracking autonomously.

When the lens actively turns toward the user and reacts to external movements, the phone gains a strange "bodily presence." To exaggerate slightly, it no longer feels like a quiet screen on a table—you might wonder, "Is this thing alive?"

(Image Source: Leikeji)

According to Honor's plans, the Robot Phone will introduce agent capabilities like perception, decision-making, and execution. Combined with its proactive environmental awareness, it will judge when to act based on user habits and emotional states.

Incidentally, this product is now available for pre-order across all channels, with an expected release in August this year.

From passively understanding user commands to actively interpreting the real world, it's clear that "proactive intelligence" has become an industry consensus.

However, the line between proactive service and meddling is often a single misjudgment away.

Promptly reminding you to board a flight feels thoughtful; unauthorized rescheduling feels invasive. Helping you track shots feels smart; observing your every move 24/7 would likely draw backlash. Questions about when it can watch, what information it can retain, and whether users can fully disable it will become more critical than how flexibly the camera rotates.

Since everyone's talking about Agents, the next question is: what will enable them to get things done?

Currently, the simplest approach is to have AI mimic human interface operations: understanding the screen via screenshots, locating buttons through visual recognition, and simulating user clicks, swipes, and inputs.

Doubao's mobile assistant gained fame through this GUI-based approach, as did the recently tested Coolpad Xiaofangkuai.

(Image Source: Leikeji)

Its advantage lies in its versatility—AI can theoretically interact with any app a human can operate, even outdated software.

The drawbacks are obvious: move a button, and the Agent might get lost; a sudden popup can derail it; encounter splash screens or privacy agreements, and artificial intelligence quickly degrades into artificial stupidity.

Moreover, this approach lacks security, once sparking widespread industry discussion about data safety.

Thus, at this year's WAIC, a new approach emerged: GUI Agents handle unadapted ordinary apps, while payments, social, and lifestyle services use A2A or MCP standard interfaces for direct collaboration between agents.

OPPO's Xiao Bu and Alipay's Abao exemplify this trend.

(Image Source: Alibaba)

When a user asks Xiao Bu to buy tickets, order food, pay bills, or handle civic services, Xiao Bu perceives the scenario, understands the intent, and distributes tasks, while Abao calls on nearly 200 Alipay services to fulfill them. Account details, preferences, and payment information remain within each party's trusted system.

Similarly, WeChat is advancing comparable collaborations with multiple smartphone manufacturers. When a phone assistant receives a request to send a message or make a WeChat call, it can hand structured tasks to Xiaowei for execution by WeChat itself, rather than letting a GUI Agent rummage through your chat list.

In my view, only by handling high-frequency, sensitive tasks well through A2A and MCP, while using GUI as a fallback for uncovered scenarios, can Agents truly transition from demos to complex real-world environments.

However, managing GUI, on-device models, cloud models, third-party agents, user memory, and system permissions simultaneously makes the traditional "smartphone assistant + Android" approach inadequate.

This is why StepFun introduced Step AOS, an agent-native operating system.

(Image Source: Leikeji)

In previous launches and interviews, Leikeji noted that StepFun identifies three barriers in existing agents: memories can't be shared across apps, single models struggle with complex decisions, and standalone apps lack access to system and other services.

Step AOS aims to reorganize these fragmented capabilities.

At the system level, Step AOS elevates Agents to a near-user system-level role, uniformly scheduling CPU, GPU, and NPU resources. It establishes cross-app, cross-device semantic memory, breaks down system capabilities like communications and files into atomic services agents can invoke, and grants permissions as needed for smoother cross-app service orchestration.

According to StepFun's vision, users won't need to say, "Open XX Maps" or articulate specific demands—they'll simply state their destination and preferences, with AgentOS planning the rest.

(Image Source: StepFun)

In an interview, StepFun Chairman Yin Qi summarized this as shifting from "process interaction" to "outcome interaction."

Coincidentally, Honor also officially upgraded MagicOS into the industry's first partner-type multimodal agent operating system, Agentic OS, adopting the next-gen Magic agent large model matrix co-developed with Alibaba. Xiaomi has similarly teased that its next-gen system will evolve into AgentOS.

Nowadays, looking at the entire industry, there is a subtle sense of mutual checks and balances among manufacturers. Mobile phone manufacturers hold the keys to systems and hardware, internet platforms dominate payments, social interactions, and lifestyle services, while large model companies wield understanding, memory, and planning capabilities.

Whoever first understands user needs and can integrate more services stands a better chance of becoming the gateway to the next generation of user experience.

After two days at the expo, Xiao Lei indeed gained a wealth of insights.

This conference not only showcased the new progress made by domestic manufacturers in on-device models, Agent smartphones, and system-level AI but, more importantly, I could clearly sense that the way mobile phone manufacturers tell their stories has changed.

The contemporary smartphone market is truly facing challenges. Based on IDC data, smartphone shipments in China during the second quarter of 2026 witnessed a 4.3% year-on-year decrease, with a 4.2% decline in the first half of the year. On a global scale, IDC anticipates a 13.9% drop in shipments this year, which would represent the largest annual decline ever recorded.

The issue isn't solely a dearth of product innovation; more significantly, the prices of storage and components are on the rise. High-capacity memory has emerged as a major contributor to costs, with storage expenses alone potentially increasing by $100 to $150 for certain flagship configurations. Manufacturers are left with little choice but to either raise prices, downgrade specifications, or reduce the number of mid-range and low-end models. When consumers are confronted with these prices, they naturally become even more reluctant to upgrade their devices.

(Image credit: Leikeji)

The mobile phone industry is in need of a fresh narrative, and Agent technology has stepped in to fill the void.

Consequently, we observe that mobile phones showcased at this year's WAIC are no longer fixated on displaying a handful of AI features. Some have begun performing tasks on behalf of users, others have developed dynamic 'eyes' to comprehend their surroundings, and some have even revamped system scheduling to ensure seamless collaboration among models, applications, and services.

In the past, our primary concern was the capabilities of our phones; now, the more pertinent question is, what can they do for us?

In this new competitive landscape, success will favor those who can minimize Agent malfunctions and deliver tangible results, as well as those who can effectively handle permissions and safeguard privacy.

WAIC 2026, under the theme 'Intelligent Partners · Co-creating the Future,' is in full swing!

The AI narrative has transitioned from focusing on accumulating model parameters to implementing Agent-driven productivity. Heterogeneous collaboration and photonic computing continue to push the boundaries of computation. Embodied intelligence is accelerating applications, with robots making their way into homes and factories, transforming physical AI into a tangible reality.

The Leikeji WAIC exploration team has arrived in Shanghai to witness the pinnacle of AI industrialization for the year. Stay tuned for more updates!

AI Doubao Agent StepFun Honor

Source: Leikeji

The images featured in this article are sourced from the 123RF licensed image library. Source: Leikeji