What is Qwen Intelligence
Qwen Intelligence is a full-stack AI phone solution launched by Alibaba's Tongyi Qianwen team, providing phone manufacturers with models, platforms, and scenario-based solutions. Qwen Intelligence includes three major Agents: Mobile Planner is responsible for task planning and decision-making, Mobile-Use executes phone operations, and Mobile Creative is oriented toward imaging creation, all achieving industry-leading SOTA performance. The solution has already been implemented in collaboration with Honor, achieving a comprehensive task accuracy of 91.8%, enabling phones to move from passive response to proactive service.
Main Features of Qwen Intelligence
-
Mobile Planner Agent (Intelligent Planning): As the phone's "brain", it understands complex needs, breaks down tasks, orchestrates tools, and dynamically adjusts, supporting long-horizon task delivery.
-
Mobile-Use Agent (Phone Operations): Adopts an "API-first + GUI fallback" model to execute cross-app operations, can find information, fill out forms, and has three layers of security and privacy controls.
-
Mobile Creative Agent (Imaging Creation): Automatically breaks down colloquial requirements into workflows, directly produces usable images from a single sentence, with first image generation taking only 3 seconds.
-
Dual-Layer Memory System: Working memory evolves with tasks, while long-term memory accumulates user profiles, understanding users better the more it is used.
-
Full-Stack Platform Capabilities: Modular and customizable, supporting independent model deployment, unified Skill governance, one-stop operations and maintenance, and automated evaluation.
-
Open Evaluation System: Opens four evaluation sets, covering full-chain assessment of planning, execution, real devices, and security.
How to Use Qwen Intelligence
- Visit the Official Website: Visit the Qwen Intelligence official website (qwenintelligence.com) to learn more about the solution.
- Enter the Qianwen AI Platform: Visit the Qwen Intelligence section of the Qianwen AI platformhttps://www.qianwenai.com/agents/qwenintelligence在线体验。
- Give Instructions in Natural Language: Directly state your needs (such as "help me arrange a business trip to Shanghai on Monday"), and the Planner Agent will automatically plan and break down the task.
- Authorize and Confirm Operations: When operations involve information retrieval, form submission, etc., the Mobile-Use Agent will request authorization, and key decisions require user confirmation before execution.
- Generate Creations with One Sentence: In creation scenarios, describe your needs in colloquial language, and the Creative Agent will automatically break down the workflow and directly produce the finished image.
Core Advantages of Qwen Intelligence
-
Full-Stack Solution: Focused on providing phone manufacturers with a one-stop AI foundation of models + platform + scenario solutions.
-
Deep Scenario Optimization: Based on the Qianwen large model and specially optimized for phone scenarios, with stronger capabilities and higher cost-effectiveness.
-
Industry SOTA Performance: The three major Agents all reach the first tier or first place in mainstream evaluations of planning, operations, and creation.
-
Extreme Cost-Effectiveness: The Planner Agent costs only about $2.41 per thousand task calls, a fraction of leading models.
-
Efficient Execution: End-to-end task success rate of 90%, task completion time reduced by 28.8%, and costs reduced by 20%.
-
Safe and Reliable: Three layers of security control, with a security score 23 percentage points higher than leading solutions, and co-building security evaluation with Fudan's Baize team.
-
Understands You Better the More You Use It: A dual-layer memory system builds user profiles, continuously personalizing throughout long-term collaboration.
Comparison of Qwen Intelligence with similar competitors
| Comparison Dimension | Qwen Intelligence | Apple Intelligence(Siri AI) |
|---|---|---|
| Product Positioning | Full-stack AI solution for smartphone manufacturers (B2B foundation) | Apple develops and uses it in-house, limited to its own device ecosystem (B2C closed loop) |
| Architecture Model | Three-agent division of labor across planning + operation + creativity, with models × Harness co-evolving | On-device self-developed models + Google Gemini-driven, deeply integrated into the system |
| Task Execution | "API-first + GUI fallback", 90% end-to-end success rate, 59.5 seconds completion time | Supports executing complex tasks across apps, but Siri AI features have been delayed multiple times and will only launch in Beta in fall 2026 |
| Memory Capabilities | Dual-layer memory: working memory + long-term user profile, understanding the user better the more it is used | Personal context understanding (emails, photos, messages), relying on iCloud sync |
| Creative Capabilities | Creative Agent generates images directly from a single sentence, first image in 3 seconds, reaching the industry's top tier | Image Playground generates images, supports writing tools, capabilities are relatively basic |
| Cost | About $2.41 per thousand tasks, a fraction of leading models | Not billed separately, but limited to high-priced new models (iPhone 15 Pro and above) |
| Degree of Openness | Four open evaluation sets, modular and customizable, already implemented with Honor (91.8% accuracy) | Closed ecosystem, only developer APIs are open, Siri AI is temporarily unavailable in mainland China |
Application Scenarios of Qwen Intelligence
-
Travel Arrangements: With a single sentence, "Help me arrange a business trip to Shanghai on Monday," it automatically compares flight and high-speed rail prices, books the hotel you stayed at last time, and sets itinerary reminders.
-
Cross-App Information Filling: Understands fragmented clues to retrieve the corresponding images, automatically pulls information to complete forms (such as reimbursement forms and registration forms), and asks the user to confirm before submission.
-
Long-Horizon Task Execution: Plans complex tasks with 100+ steps (such as planning an event), gathers fragmented knowledge, and gradually advances delivery.
-
One-Sentence Image Creation: Using a spoken description, "Make a clearly laid-out retro diary," it automatically breaks down the workflow and directly produces a finished image in 3 seconds, with no need for professional prompts.
-
Browser and Cross-Device Operations: Capabilities extend to browser scenarios and multi-device collaboration, achieving an end-to-end service closed loop (the Honor solution has already reached a 90% closed-loop rate).