MetaEra Launches the World’s First AI-Powered Smartphone with Alipay and Meituan Integration

Shanghai-based AI unicorn StepFun has launched three new products: the world’s first agent-native operating system, Step AOS; the personal agent Amoo; and the AI terminal brand STEPX, alongside the debut of its first agent-powered smartphone, STEPX Neo. The smartphone features StepFun’s proprietary Step AOS system, enabling users to express intentions in natural language, allowing agents to autonomously execute cross-app tasks. The system supports long-term memory and edge-cloud collaboration, with auditable and reversible operations. Initial ecosystem partnerships include leading apps such as Alipay, Meituan, AutoNavi, Didi, JD.com, and Baidu. Over the next 100 days, StepFun will continue building its agent ecosystem, with its flagship reasoning model, Step Pro, boasting over 1.5 trillion parameters.

Article author and source: Zhixidong

100 days from now, more updates on this phone will be disclosed.

Alipay

Unfunded

Third-party payment service platform

Zhixidongxi reported on July 13 that, just moments ago, Shanghai-based large model unicorn Jieyue Xingchen unveiled three new products at its launch event: the world’s first agentic-native operating system, Step AOS (Step Agentic-native OS); the personal agent Amoo; and the large model-native AI terminal brand STEPX.

The world's first large model-native intelligent agent phone, STEPX Neo, which was highly anticipated by the public, was also unveiled at today's event.

STEPX Neo is powered by StepStar’s proprietary agent-native operating system, Step AOS, and comes preloaded with the next-generation system-level personal agent, Amoo. Users no longer need to manually open individual apps—simply state your intent as if chatting with an assistant, and the agent will autonomously understand, plan, and orchestrate tasks across applications to complete them.

Step AOS also features long-term memory and the ability to collaboratively process complex tasks between edge and cloud, with all operations being auditable and reversible to ensure security and control.

STEPX Neo's positioning is similar to the much-discussed "Doubao Phone" in the industry—it also needs to address ecosystem challenges and encourage apps to grant permissions to agents. Currently, StepCraft has established ecosystem partnerships with multiple leading apps, including Alipay, Meituan, AutoNavi, Didi, JD.com, and Baidu.

Jieyue Xingchen Chairman Yin Qi said that today’s launch event is only the first half; over the next 100 days, they will collaborate with users to build an agent ecosystem, including various agents and skills. At the next launch event after 100 days, more agents will be unveiled.

During the post-event media Q&A, Yin Qi and Ni Jiayue, CEO of Step Terminal, shared Step Infinity’s first-phase goal for AI terminals: not merely pursuing shipment volume, but ensuring scale, with the aim of getting early users on board as quickly as possible.

Yin Qi said that DouBao Phone represents a significant exploration for the entire industry, truly bringing the concept of agent phones into the public eye. Jiepoin星辰 has unique perspectives on the definition of agent phones and operating systems, and he looks forward to the emergence of more new agent phone brands.

In his vision, future intelligent agents may no longer be phones—you’ll hardly use them to make calls, and longer interaction times won’t be better; instead, they might accomplish tasks for you with just a simple interaction.

Yin Qi revealed that StepFun had long debated whether to enter the hardware market themselves and had heard many voices discouraging them, but they ultimately decided to launch STEPX, an AI-native intelligent agent terminal brand.

STEPX's products are AI-native hardware. Stepwise defines such hardware as having zero value without AI. Additionally, this type of hardware must achieve an integrated, synergistic development of models, software, and hardware.

STEPX's first product, STEPX Neo, debuted today, and Yin Qi emphasized that this is not an official "launch."

Why did Stepwise choose a smartphone as the first agent terminal? Stepwise believes that, in the future, the primary device for agents should still be a portable device capable of frequent interaction with users.

Second, the screen remains the primary means of interaction between users and machines, so agent devices also require a screen. Finally, this device needs to have sufficient on-device computing power.

The combination of these three points ultimately comes back to the form of a smartphone.

The launch event did not detail specific specifications of the device, but instead devoted significant attention to its ecosystem. After the hardware product’s system and model were finalized, StepFun partnered with several major Chinese internet companies to build its ecosystem. The initial ecosystem partners include Alipay, Meituan, AutoNavi, Tongcheng, CapCut, JD.com, Didi, Baidu, Weibo, and WPS.

Yin Qi said that many people believe there is a natural divide or contradiction between agent interfaces and apps, but this view is fundamentally wrong. He believes that, from software in the PC era, websites in the internet era, to apps in the mobile internet era, application formats have continuously evolved—applications will never disappear; instead, they will become more numerous and adopt a wider variety of forms, with intelligent agents in the AI era becoming the next generation of application carriers.

To further build its ecosystem, Jiepai has partnered with Bilibili to launch the "Rhapsody Project," inviting a thousand creators to experience the device early and develop engaging intelligent agents. Jiepai is also actively gathering user feedback through its 100-Day Co-Definition Initiative, aiming to create diverse intelligent agents based on its hardware operating system and models.

To truly bring intelligent agent hardware like STEPX Neo to market, Jiepoin星辰 needs to make further innovations at the system and agent levels. Therefore, Jiepoin星辰 has launched Step AOS, an agent-native operating system, and the next-generation personal agent, Jiepoin Amoo.

Among these, Step AOS has been restructured across three levels: system capabilities, interaction methods, and underlying infrastructure.

Memory capability has become a standard feature for many agents. Step AOS implements a dual-domain, three-step memory structure: the user domain enables the agent to remember and understand users, while the agent domain accumulates its own knowledge and experience.

This information is not simply accumulated, but rather processed in three stages: first recorded during everyday interactions, then organized, deduplicated, and consolidated in the background, and finally retrieved when needed.

In terms of decision execution, Step AOS enables seamless and flexible collaboration and division of labor between on-device and cloud models. For simpler tasks—such as setting an alarm or finding a photo—Step AOS follows the principle of “on-device whenever possible.” For highly complex, multi-step tasks, the cloud model takes responsibility.

Yin Qi believes that the more common pattern will still be a hybrid of edge and cloud, working in tandem to complete tasks—the edge providing real-time feedback and the cloud delivering deep analytical results. Meanwhile, model routing should primarily be determined by the models themselves, rather than fixed rules.

The end-cloud collaboration system is supported by Jiepai’s family of models, ranging from 300M to 400B parameters, covering both edge and cloud sides. The presentation slides also revealed that Jiepai’s flagship Step Pro inference model is set to be released, with over 1.5 trillion parameters.

Step Edge, the family of edge-based foundation models newly released by Step yesterday, is specifically designed for terminal hardware and ranks first globally among similar edge models across 29 authoritative benchmarks. Additionally, these models achieved the top ranking in China on benchmarks for GUI, agents, and terminal tasks.

Security is equally important for an agent-native operating system. Step AOS defines security across four dimensions: trustworthiness, visibility, controllability, and reversibility.

In this operating system, agents will complete tasks within a trusted execution environment; their operations are auditable and traceable, permissions are granted on-demand and revoked immediately after use, and accidental actions can be undone with a single click.

At the interaction layer, Step AOS primarily shifts from “process-based interaction” to “outcome-based interaction,” eliminating the need for users to issue step-by-step commands; instead, users simply express their task intent, and the agent completes the entire workflow, proactively delivering task plans, execution progress, and recommendations.

At the underlying hardware and software infrastructure level, Step AOS has redesigned computing power, data, and application services for agents, achieving unified scheduling of heterogeneous computing resources such as CPUs, GPUs, and NPUs through a unified computing resource pool, while also building a unified semantic data layer for processing to enable high-speed data transmission.

In terms of applications and services, Step AOS’s atomic capability engine decomposes the original system’s functions and services into atomic services suitable for agent scheduling and orchestration, and opens these capabilities to the entire ecosystem.

The ultimate goal of these designs is to make the system better suited for human-machine symbiotic usage.

Leveraging the system capabilities of Step AOS, Jiepao Xingchen has launched Jiepao Amoo, a new-generation personal agent. This is a system-level agent capable of cross-app orchestration, edge-cloud collaboration, and executing tasks across different devices.

In Qin Qi’s view, constrained by the mobile internet architecture, intelligent agents still face three walls: the memory wall, the decision wall, and the action wall.

Specifically, data fragmentation between apps prevents AI from maintaining cross-app memory, limiting its ability to deliver better services to users. Second, most agents can only access a single model, which struggles to balance efficiency with complex reasoning, constraining decision-making capabilities. More importantly, agents remain confined within app permission systems and cannot truly accomplish tasks on behalf of users.

The design of Step AOS and Amoo aims to provide solutions to the above challenges.

At this launch event, Jiepoin星辰 aimed to bridge the final mile from agents that “can think and speak” to those that “can truly get things done” through an integrated hardware-software approach, end-to-cloud collaboration, and ecosystem cooperation.

Although the hardware details revealed in this launch are limited, the pace of its ecosystem development demonstrates stronger practical advantages compared to the DouBao phone at a similar stage.

Of course, whether intelligent agent phones can truly break down data barriers between apps, continuously expand the breadth and depth of the app ecosystem, and encourage users to adopt and rely on this new interaction model will ultimately be tested by the market and time.