🔥36氪•Freshcollected in 2m
StepFun launches world's first native agentic phone
💡The first 'native agentic phone' is here—see how OS-level model integration changes the mobile AI landscape.
⚡ 30-Second TL;DR
What Changed
Step AOS operating system powers the device
Why It Matters
This launch marks a shift from 'AI-enabled' to 'AI-native' hardware, potentially changing how users interact with mobile OS environments.
What To Do Next
Explore the StepFun developer documentation to understand how to build agents compatible with the Step AOS ecosystem.
Who should care:Developers & AI Engineers
Key Points
- •Step AOS operating system powers the device
- •Built-in Amoo agent enables native AI interaction
- •Integrates model, software, and hardware as a new AI species
🧠 Deep Insight
AI-generated analysis for this event.
🔑 Enhanced Key Takeaways
- •StepFun, founded by former ByteDance AI lab head Li Lei, has transitioned from a pure model provider to a full-stack hardware integrator with this release.
- •The Amoo agent utilizes a 'System-Level' architecture that allows it to bypass traditional app sandboxing to perform cross-application tasks autonomously.
- •The device employs a proprietary 'Step-2' multimodal large model architecture optimized specifically for on-device latency and low-power execution.
- •The operating system features a 'Dynamic UI' that generates interface elements in real-time based on the agent's current task context rather than relying on static app icons.
- •The hardware includes a dedicated NPU cluster designed to handle continuous multimodal input (vision, audio, and text) without offloading to the cloud.
📊 Competitor Analysis▸ Show
| Feature | StepFun (Amoo) | Apple (Intelligence) | Samsung (Galaxy AI) |
|---|---|---|---|
| Agentic Depth | Native System-Level | App-Integrated | App-Integrated |
| UI Paradigm | Generative/Dynamic | Static/App-based | Static/App-based |
| Model Origin | Proprietary (Step-2) | Hybrid (On-device/Cloud) | Hybrid (On-device/Cloud) |
| Pricing | Mid-to-High Tier | Premium | Premium |
🛠️ Technical Deep Dive
- Architecture: Utilizes a unified transformer-based backbone that processes multimodal tokens (text, image, audio) in a single stream to reduce latency.
- Memory Management: Implements a 'Context-Aware' memory buffer that stores user interaction history locally to enable long-term personalization without cloud synchronization.
- OS Integration: Step AOS replaces the traditional Android framework layer with an agent-first middleware that intercepts system calls to allow the Amoo agent to control hardware sensors and app APIs directly.
- Power Efficiency: Employs a tiered inference strategy where simple tasks are handled by a lightweight distilled model, while complex reasoning triggers the full Step-2 model.
🔮 Future ImplicationsAI analysis grounded in cited sources
StepFun will force a shift toward 'Agent-First' OS design in the Chinese smartphone market.
The success of a native agentic OS will compel major OEMs like Xiaomi and OPPO to move beyond simple AI features to deep system-level integration.
The device will face significant regulatory scrutiny regarding data privacy and autonomous system control.
Granting an AI agent system-level access to cross-app data and hardware controls creates new security vulnerabilities that current mobile OS regulations are not equipped to handle.
⏳ Timeline
2023-08
StepFun (Jieyue Xingchen) is founded by Li Lei and a team of former ByteDance AI researchers.
2024-03
StepFun releases the Step-1 multimodal large model, marking its entry into the foundation model market.
2025-01
Company announces the development of the Step-2 model, focusing on advanced reasoning and agentic capabilities.
2026-05
StepFun begins internal testing of the Step AOS operating system on prototype hardware.
2026-07
Official launch of the world's first native agentic phone featuring Amoo and Step AOS.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪 ↗