- Chinese AI startup StepFun has detailed its STEPX smartphone running Step AOS 1.0, an operating system engineered specifically for an autonomous AI agent rather than serving as an add-on assistant.
- Priced at 3,999 yuan, the hardware features a proprietary agent named Amoo, dual interfaces, a rear MiniLED status screen, and a 12-month STEPX Token Plan offering 100 billion monthly tokens.
- Nearly 100 domestic platforms—including Alipay, Tencent, Ctrip, Didi, and Meituan—have integrated with Amoo via the Model Context Protocol (MCP) to manage end-to-end tasks directly.
As smartphone manufacturers race to integrate generative artificial intelligence into existing operating systems, Chinese AI company StepFun is pursuing an alternative architecture by repositioning the smartphone directly around an autonomous agent. Rather than requiring users to manually launch ride-hailing applications, input destination addresses, and select vehicles, StepFun’s hardware allows users to vocalize their intent while the system’s software independently executes the workflow from start to finish. StepFun told the launch audience that in the agent era AI is no longer an add-on to a smartphone but the product itself.
Step AOS 1.0 and the Ground-Up Agent Architecture
StepFun first publicly previewed the autonomous agent, its proprietary operating system, and the STEPX brand in Shanghai on July 13 without detailing pricing or a commercial release schedule. During the presentation, the company stated that conventional smartphone software was never built for autonomous task execution. While platforms such as Apple’s Siri and Google’s Gemini operate as embedded assistants inside iOS and Android, StepFun says its Step AOS 1.0 was instead built from scratch for the agent, with long-term memory, task scheduling, and security controls in the base layer.
The underlying software is characterized by StepFun as an integration of Android, Linux, and real-time operating system components. Step AOS decomposes primary mobile functions—including messaging, applications, storage files, and system settings—into discrete, modular blocks that the agent can summon directly. This workflow utilizes the Model Context Protocol (MCP), an open and shared standard designed to interface AI programs with third-party software. To preserve standard usability, the device provides users with two distinct environments: a traditional application grid and a dedicated operational space for the autonomous agent.
Hardware Integration and the Amoo Ecosystem
The device’s built-in agent, named Amoo, relies on a bifurcated memory system. One partition stores user-specific data, behavioral patterns, and personal preferences, while the other maintains learned routines and contextual knowledge for executing multi-step tasks. Operational workloads are dynamically allocated: routine and localized requests, such as configuring alarms, run on-device via a compact internal model (operating offline when necessary), whereas complex requests spanning multiple services are directed to a cloud-based foundational model. Physical hardware features include a dedicated secondary MiniLED display located on the rear of the phone, which provides real-time visual feedback on the actions Amoo is performing.
To avoid the screen-reading and simulated-click mechanics deployed by platforms like Doubao, StepFun has formed service integration agreements with nearly 100 enterprise partners and services. Major companies including Alipay, Tencent, Ctrip, Didi, and Meituan have joined the Amoo ecosystem. These direct application-level integrations allow actions such as digital payments, itinerary bookings, and multimedia creation to remain securely completed on their respective native platforms.
Proprietary Foundation Models and Market Entry
StepFun was established in 2023 by Jiang Daxin, a former vice president at Microsoft, with a focus on developing in-house large language models (LLMs). The company’s flagship model, Step 5 Preview, debuted in September. The model features 600 billion total parameters, with 27 billion active parameters engaged during dynamic processing, and supports an extensive context window handling up to 1 million tokens simultaneously. StepFun previously released Step 3.5 Flash, an open-source model distributed under an Apache 2.0 license.
The STEPX smartphone enters the consumer market at a retail price of 3,999 yuan. The purchase price includes 12 months of the STEPX Token Plan, granting users an allocation of 100 billion tokens per month to power cloud-based agent computations. StepFun has not yet disclosed the subscription costs or token pricing models that will apply once the initial one-year plan concludes.
Why This Matters
The debut of the STEPX phone reflects a major architectural divergence in the global consumer electronics sector. While mainstream technology giants have focused on layering AI capabilities over mature, legacy mobile operating systems, StepFun’s reliance on Step AOS 1.0 represents an attempt to bypass existing user-interface paradigms altogether through native Model Context Protocol (MCP) links. Backed by founder Jiang Daxin’s engineering pedigree, StepFun’s integration of commercial giants like Tencent, Didi, and Alipay signals a shift toward collaborative ecosystem APIs over brittle UI automation. Furthermore, incorporating recurring token allowances into retail hardware establishes a potential blueprint for AI-hardware monetization, making post-first-year subscription rates a critical metric to watch.
Frequently Asked Questions
How does Step AOS 1.0 differ from iOS or Android?
Unlike standard mobile operating systems where voice assistants function as secondary software layers, Step AOS 1.0 combines Android, Linux, and real-time operating system components to embed long-term memory, scheduling, and security directly into the base system. It breaks applications, messaging, and system settings into modular blocks that its AI agent, Amoo, can call directly via the Model Context Protocol.
Which third-party services support the Amoo agent?
Nearly 100 companies and products have integrated with Amoo, including prominent Chinese digital platforms such as Alipay, Tencent, Ctrip, Didi, and Meituan. Through direct integrations, transactions such as payments, ride bookings, and travel arrangements are routed directly to the native platforms rather than relying on automated screen clicks.
What specifications and service plans are included with the STEPX phone?
The STEPX phone retails for 3,999 yuan, which includes a 12-month STEPX Token Plan providing 100 billion computational tokens per month for cloud processing alongside its on-device models. The phone also incorporates a rear MiniLED display to illustrate agent tasks. StepFun has not yet announced post-year-one pricing details for the token subscription.
