China Unveils L1-L4 AI Device Intelligence Standard

China Launches First National AI Device Intelligence Rating System

China has officially introduced a national standard that classifies AI-powered consumer devices into 4 intelligence levels, ranging from basic responsiveness to full collaborative autonomy. The 'AI Terminal Intelligence Grading' standard (GB/Z 177—2026), published on May 8 by the Ministry of Industry and Information Technology (MIIT) alongside China's market regulation and commerce agencies, covers smartphones, PCs, smart TVs, AR/VR glasses, automotive cockpits, smart speakers, and earphones.

The move represents the world's first government-backed attempt to create a universal grading system for on-device AI capabilities — a framework that could reshape how consumers evaluate and compare smart devices across categories. Major Chinese tech firms including Huawei, Xiaomi, Lenovo, OPPO, vivo, Honor, and iFlytek all participated in drafting the standard.

Key Takeaways

• 4 intelligence tiers: L1 (Response), L2 (Tool), L3 (Assistant), and L4 (Collaborative), with increasing levels of device autonomy
• 7 product categories covered in the first batch: smartphones, PCs, TVs, smart glasses, automotive cockpits, speakers, and earphones
• '2+N' architecture: 2 foundational documents define the framework, with 'N' category-specific standards for each device type
• L4 remains aspirational: The highest tier will be further defined as the industry matures
• Major Chinese OEMs co-authored the standard, signaling broad industry alignment
• Additional device categories will be added in future revisions

Breaking Down the L1-L4 Intelligence Tiers

The grading framework establishes a clear hierarchy of device intelligence that mirrors, in some ways, the well-known autonomous driving levels used by SAE International. Each level represents a meaningful step up in how independently a device can understand, reason, and act on behalf of its user.

L1 — Response Level represents the most basic tier. Devices at this level can respond to direct user commands but lack contextual understanding. Think of a smart speaker that plays a song when asked but cannot anticipate preferences or chain tasks together.

L2 — Tool Level describes devices that function as capable utilities. These gadgets can execute more complex instructions, handle multi-step workflows, and provide structured outputs — similar to how current AI assistants process queries and return formatted results.

L3 — Assistant Level marks a significant leap. Devices at this tier proactively assist users, understanding context, learning preferences over time, and offering suggestions without explicit prompting. This is where most flagship smartphones with integrated AI features — like Apple Intelligence, Google Gemini on Pixel, or Samsung Galaxy AI — are aiming today.

L4 — Collaborative Level represents the aspirational peak. At this stage, devices act as true collaborative partners, autonomously managing complex tasks, coordinating across ecosystems, and adapting in real time. The standard acknowledges that L4 remains forward-looking and will be refined as technology evolves.

Why a National Standard Matters for Global Tech

China's decision to formalize AI device intelligence grading carries implications far beyond its domestic market. As the world's largest smartphone market and a major hub for consumer electronics manufacturing, China's standards frequently influence global product development cycles.

For Western companies like Apple, Google, Microsoft, and Samsung that sell devices in China, this standard could become a de facto benchmark they need to meet — or at least reference — when marketing AI capabilities to Chinese consumers. It also creates pressure on other governments and standards bodies to develop comparable frameworks.

The European Union, which has taken a regulatory-first approach to AI through the EU AI Act, has focused primarily on risk classification for AI systems rather than device-level intelligence grading. China's standard fills a different niche: it is less about safety guardrails and more about establishing a common language for measuring what 'smart' actually means in consumer hardware.

This distinction matters. While Brussels asks 'how risky is this AI?', Beijing is asking 'how smart is this device?' — and providing a standardized answer that manufacturers and consumers can reference.

The '2+N' Architecture Explained

The standard adopts a modular '2+N' structure designed for scalability. The '2' refers to 2 foundational documents that apply universally across all device types.

Part 1: Reference Framework defines core concepts — what qualifies as an AI terminal, how intelligence is conceptualized, and the overall classification methodology. Part 2: General Requirements specifies the testing methods and evaluation criteria used to assign an intelligence grade to any device.

These 2 documents serve as the bedrock for all subsequent category-specific standards, ensuring consistency whether the device being evaluated is a pair of smart earbuds or an automotive cockpit system.

The 'N' component encompasses product-specific standards tailored to each device category. The first batch covers 7 categories:

• Mobile terminals (smartphones and tablets)
• Microcomputers (laptops and desktops)
• Televisions
• Smart glasses (AR/VR/MR devices)
• Automotive cockpits
• Smart speakers
• Earphones and headphones

Additional categories — potentially including wearables, home appliances, and robotics — will be developed in subsequent phases.

Industry Players Shaping the Standard

The list of companies involved in drafting the standard reads like a who's-who of China's tech ecosystem. Huawei and Xiaomi, both aggressively pushing on-device AI capabilities in their smartphone lineups, are co-authors. So are OPPO and vivo, which have been integrating large language models directly into their mobile operating systems.

Lenovo, the world's largest PC maker by shipments, brings the computing perspective. Honor, Huawei's former sub-brand now operating independently, has been marketing AI-enhanced photography and productivity features across its device portfolio. iFlytek, China's leading voice AI company, contributes deep expertise in natural language processing and speech recognition.

Notably absent from the publicly listed drafting companies are major Western firms. This raises questions about whether the standard will adequately account for the AI architectures and capabilities prioritized by companies like Apple, Google, or Qualcomm — all of which have significant AI-on-device strategies.

For context, Apple's Intelligence platform, Google's Gemini Nano, and Qualcomm's AI Engine represent the primary on-device AI frameworks used in Western markets. How these map to China's L1-L4 grading will be a critical question for multinational device makers.

What This Means for Consumers and Businesses

For everyday consumers, this standard promises something deceptively simple: clarity. Today, nearly every device manufacturer claims AI capabilities, but there is no universal way to compare what 'AI-powered' actually means across brands and product categories.

A standardized grading system could function like an energy efficiency rating — giving buyers a quick, reliable indicator of a device's AI sophistication. Imagine walking into a store and seeing an 'L3 AI' label on a smartphone, immediately understanding it offers proactive, context-aware assistance rather than basic voice commands.

For businesses and developers, the implications are more strategic:

• Product positioning: Manufacturers will compete not just on specs but on intelligence tier placement
• Software optimization: App developers may need to target specific intelligence levels, similar to how apps target different OS versions
• Supply chain decisions: Component suppliers (chipmakers, sensor manufacturers) will align roadmaps with tier requirements
• Marketing standardization: Vague AI marketing claims could give way to verifiable tier classifications

Looking Ahead: Global Ripple Effects

The full text of the standard has not yet been published on China's National Standards Information Public Service Platform, meaning the detailed technical criteria remain under wraps. However, the framework's release signals that China intends to move quickly on implementation.

Several key developments to watch in the coming months include whether international standards bodies like ISO or IEC pursue harmonized frameworks, how Western regulators respond, and whether major global OEMs voluntarily adopt China's grading in their product marketing.

The timing is significant. The global on-device AI race is intensifying, with Apple, Google, Samsung, and Qualcomm all making major investments in edge AI processing. China's standard arrives at precisely the moment when the industry needs — but has so far lacked — a common vocabulary for device intelligence.

Whether this particular framework becomes the global template or merely the first of several competing systems, one thing is clear: the era of undefined, unverified AI marketing claims for consumer devices may be drawing to a close. The L1-to-L4 ladder gives the industry a structured way to measure progress — and gives consumers a reason to demand proof behind the promises.