Alibaba unveils Qwen Intelligence for agentic AI smartphones, HONOR is first partner

Alibaba has unveiled Qwen Intelligence, a full-stack agentic AI solution designed to help smartphone manufacturers integrate advanced AI agents into their devices, with HONOR becoming its first partner. The two companies are co-developing domain-specific models and solutions for next-generation AI smartphones.

Steven Hoi Head of Qwen Mobile AI at Token Foundry and Vice President of Alibaba Token Hub ATH launched Qwen Intelligence at Apsara Conference

Why it matters: Qwen Intelligence goes beyond putting a general-purpose AI model inside a smartphone. Alibaba is providing manufacturers with a complete stack covering mobile-optimized foundation models, an agent platform, tools, and ready-to-use AI agents that can perform tasks across applications.

The partnership with HONOR also points toward smartphones becoming more proactive, with AI systems capable of planning and executing multi-step tasks rather than simply responding to individual prompts.

The details: Alibaba introduced Qwen Intelligence at the Apsara Conference in Hangzhou. The platform is designed specifically for mobile scenarios and can be adopted as a complete solution or integrated selectively, allowing smartphone manufacturers to use individual components within their own systems.

“Agentic smartphones are the primary gateway bridging personal AI and the physical world. Qwen Intelligence integrates mobile-optimized Qwen models with a custom-engineered harness, empowering phone makers to deliver seamless, personal intelligence to their users,” said Steven Hoi, Head of Qwen Mobile AI at Token Foundry and Vice President of Alibaba Token Hub.

The first phase of Qwen Intelligence includes three ready-to-use agents designed to help smartphones understand users, plan tasks, and perform actions.

Mobile Planner Agent serves as the planning layer, handling task decomposition, tool orchestration, and dynamic adjustments. For example, a user could ask the phone to arrange a business trip to Shanghai, with the agent breaking the request into tasks such as booking flights, reserving a hotel, adding calendar events, and preparing an itinerary. If a flight is canceled, its memory and proactive services can assist with rebooking and finding alternatives.

Mobile-Use Agent is designed to perform actions across smartphone applications using a hybrid API-first and graphical user interface fallback approach. Alibaba says the system incorporates security boundaries and privacy protections to help ensure tasks are executed safely.

Mobile Creative Agent is a lightweight model optimized for smartphone-based content creation, including image generation and editing. Additional services in development include personalized memory and proactive assistance, multimodal interaction, and specialized vertical agents developed with ecosystem partners.

Qwen Intelligence uses a three-layer architecture. At the foundation is a Qwen-based model optimized for mobile scenarios, which Alibaba says can provide stronger on-device capabilities and lower costs compared with adapting a general-purpose model for smartphones. The middle platform layer is modular and supports independent model deployment, AI harness customization, unified tool governance, and automated evaluation. Device-cloud coordination also allows manufacturers to balance performance, latency, and cost.

At the top are premium agents designed for specific domains and real-world use cases. Smartphone manufacturers can deploy the complete Qwen Intelligence stack or select individual components, from foundation models to specific agents, and integrate them into their existing software systems.

Alibaba has also developed an evaluation system to measure how AI agents perform in real-world mobile scenarios. MobilePA-Bench evaluates an agent’s ability to plan and execute complex tasks from beginning to end, covering more than 1,000 real-world scenarios and more than 200 commonly used mobile tools. Additional benchmarks evaluate long-horizon and cross-application tasks in cloud-phone environments, as well as tasks performed on physical devices that reflect everyday smartphone usage.

Qwen Intelligence is already being applied through Alibaba’s collaboration with HONOR. The companies are developing domain-specific models and solutions that combine Qwen Intelligence with HONOR’s MagicOS operating system, with the goal of moving smartphone AI from responding to individual commands toward proactive services capable of coordinating multiple actions.

According to Alibaba, HONOR AI smartphones using the architecture can achieve task accuracy of up to 91.8% and orchestrate complex workflows involving more than 100 steps.

The bigger picture: The collaboration comes as smartphone makers increasingly position AI agents as a new layer of interaction between users and their devices. Instead of requiring users to manually open and switch between applications, agentic systems are designed to understand a goal and coordinate multiple actions to complete it.

HONOR’s upcoming Magic9 Series, scheduled to launch on September 28, and the HONOR Robot Phone will be among the first devices incorporating Qwen Intelligence.

What’s next: The initial HONOR devices will provide an early look at how Alibaba’s mobile-optimized Qwen models and agent framework can be integrated into a smartphone operating system, particularly for complex tasks involving multiple applications and services.

The Bottom Line: Qwen Intelligence gives smartphone manufacturers a complete framework for developing agentic AI experiences, combining mobile-optimized models, an orchestration platform, and ready-to-use agents. Its first major smartphone partnership with HONOR will bring the technology to consumer devices through MagicOS, starting with the Magic9 Series and HONOR Robot Phone.


Discover more from WalasTech

Subscribe to get the latest posts sent to your email.

Get the latest from WALASTECH directly on your Google feed.

Add as a preferred source on Google
Carl walked away from a corporate marketing career to build WalasTech from the ground up—now he writes no-fluff tech stories as its Founder and Editor-in-Chief. When news breaks, he’s already typing. Got a tip? Hit him up at [email protected].