The phone market has welcomed a new AI player.

On October 13, Step Terminal officially launched its first large-model-native agent phone, the STEPX Neo, with a pioneer edition priced at 3,999 yuan. It will officially go on sale on October 16, available simultaneously on platforms such as JD.com, Tmall, and Douyin.

Unlike traditional phone makers, which mainly compete around hardware configurations such as chips, imaging, and battery life, Step has placed the personal agent at the core of the product.

The newly released STEPX Neo is equipped with the agent-native operating system Step AOS 1.0, as well as a new-generation personal agent, Step Amoo, in an attempt to change the past usage pattern in which users open different apps one by one and complete tasks step by step, allowing AI to understand user needs and call different applications to complete tasks.

This is an attempt by Step to move further from large-model research and development toward consumer terminals.

To make the personal agent truly participate in daily life, Step redesigned everything from the phone hardware and operating system to model capabilities.

Among these, Step AOS 1.0 integrates long-term memory, task scheduling, and security mechanisms into the underlying system. Its "dual-domain memory" system, on the one hand, records users' personal information, past experiences, and usage preferences, and on the other hand, accumulates the knowledge and experience the agent gains while executing tasks.

At the launch event, Step Xingchen Chairman Yin Qi said that a good personal agent needs to have global memory, understand user intent, and be able to execute tasks safely. Around these requirements, Step AOS 1.0 builds system capabilities in areas such as memory, decision-making and execution, and security and controllability.

He believes that in the future, the memory and storage capacity of phones will further expand, because phones not only need to record user information, but also need to store a large amount of agent-related information.

At the same time, how much a personal agent can accomplish also depends on how much trust users are willing to give it.

To this end, the STEPX Neo introduces mechanisms such as privacy encryption, permission management, execution process tracing, and rollback of risky operations, hoping that while users hand tasks to AI, they still retain ultimate control.

At the model level, the STEPX Neo adopts a device-cloud collaborative approach. Simple tasks such as adjusting system settings and creating alarms can be quickly executed by the on-device model, and some operations can also be completed in a no-network environment. For complex tasks such as information queries and multi-app collaboration, the cloud-side model handles planning and decomposition, which is then handed to the device side for execution.

In terms of hardware, the new phone is equipped with a six-microphone array, a low-power audio sensing chip, and an independent MiniLED dot-matrix rear display, which respectively handle functions such as the agent's voice perception and status feedback.

However, for an agent phone to truly handle things for users, models and an operating system alone are not enough. The key also lies in whether it can mobilize enough third-party applications and services.

At this launch event, Step also announced its latest ecosystem progress.

At present, Step Amoo has established cooperation with nearly 100 companies or products, including Alipay, Tencent, Ctrip, Didi, Jianying, Amap, and Meituan, covering scenarios such as local life, travel, office work, and content creation.

Among them, products under Tencent such as WorkBuddy, WeCom, and WeChat Reading became the first batch of Tencent ecosystem applications to be integrated.

Taking the office scenario as an example, when a user asks Amoo, "Help me organize last week's progress," Amoo can combine the user's past work records, project background, and personal habits to supplement the information required for the task, and then hand the organized task instructions to WorkBuddy for execution.

WorkBuddy then completes operations such as material organization and document generation in a cloud environment, and returns the task status and final output to Amoo. Users can directly view the results on their phones, or jump to WorkBuddy to learn about the specific execution process.

Tencent WorkBuddy product expert Wang Shengjie said at the launch event that the core of the cooperation between the two sides is to let users simply express their goals without needing to specify which application to use, with the subsequent tasks completed by the underlying services.

According to the plan announced by both sides, WorkBuddy can run as an independent application, can also connect to the user's PC, and can further integrate into Amoo in the form of an office skills expert.

In addition to Tencent, Alipay has also become an important ecosystem partner of Step. The two sides are cooperating around AI payments and life services, and Step Amoo can call relevant Alipay service capabilities after obtaining user authorization.

In content creation, Step Amoo can engage in cross-agent collaboration with the Jianying assistant to complete tasks such as video generation and editing modifications.

Step also launched the AI terminal-native development tool Step Studio, allowing developers to use natural language to generate AI apps, games, agents, and Skills, and to call hardware capabilities such as the phone camera, sensors, and rear display.

In terms of business model, the STEPX Neo pioneer edition adopts so-called "software-hardware-model" full-stack pricing, integrating the phone hardware, operating system, model, and development tools into the price of 3,999 yuan. Purchasing users can also receive a model usage quota of 10 billion tokens per month for 12 consecutive months.

From this launch, it can be seen that Step hopes to establish a personal agent entry point through its own terminal, and then use external ecosystems to fill in actual task execution capabilities.

For an AI company newly entering the phone market, this path must face competition from mature phone makers in hardware, channels, and user scale, while also needing to continuously push different applications to open up their capabilities.

As personal agents gradually enter phones, calling life service apps and completing tasks across applications is becoming a new direction of exploration.

But for consumers, compared with opening apps themselves to complete operations, how much time AI can save, how much efficiency it can improve, and how much new usage value it can bring still remain to be verified in practice. This will also become the key to whether agent phones can truly move consumers.