GB/Z 177.2-2026 Intelligence grading of artificial intelligence terminal—Part 2: General requirements English, Anglais, Englisch, Inglés, えいご
This is a draft translation for reference among interesting stakeholders. The finalized translation (passing through draft translation, self-check, revision and verification) will be delivered upon being ordered.
ICS
CCS
National Standard of the People's Republic of China
GB/Z 177.2-2026
Intelligence grading of artificial intelligence terminal - Part 2: General requirements
人工智能终端智能化分级 第2部分：总体要求
Issue date: 2026-03-31 Implementation date: 2026-10-01
Issued by the General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China
the Standardization Administration of the People's Republic of China
Contents
Foreword
Introduction
1 Scope
2 Normative References
3 Terms and Definitions
4 Abbreviations
5 Overview
6 Classification Levels
6.1 L1 Response Level
6.2 L2 Tool Level
6.3 L3 Assistance Level
6.4 L4 Collaboration Level
7 Capability Elements
7.1 Primary Capability Elements
7.2 Secondary Capability Elements
8 Key Capabilities
8.1 Overview
8.2 End Side Capabilities of High Computing Power Terminals
8.3 Device Cloud Collaboration Capabilities
9 Test Methods
9.1 Test Principles
9.2 Test Framework
9.3 Intelligence Level Determination
9.4 Test Procedure
Annex A (Informative) User Instructions and Intents
Annex B (Informative) Intelligence Level Scoring Model
Bibliography
Artificial intelligence terminal intelligence classification — Part 2: General requirements
1 Scope
This document specifies the classification system for the intelligence level of artificial intelligence terminals (hereinafter referred to as “terminals”), including the terminal intelligence level classification, capability elements and key capabilities for each level, and describes the test methods.
This document applies to the intelligence classification of various types of terminals, and also provides a reference for the design, development, application, selection and testing of terminals.
2 Normative References
The following document is essential for the application of this document. For dated references, only the edition cited applies. For undated references, the latest edition (including any amendments) applies.
GB/Z 177.1 Artificial intelligence terminal intelligence classification — Part 1: Reference framework
3 Terms and Definitions
For the purposes of this document, the terms and definitions given in GB/Z 177.1 and the following apply.
3.1 testing scenario
A comprehensive test context set for testing the key capabilities of a terminal in a specific application scenario.
3.2 testing task
A basic execution unit in terminal intelligence testing.
NOTE: A testing task can serve either as a specific step within a testing scenario or as an independent, atomic test item to be completed separately.
3.3 scenario context
A dynamic set of information about the user, device and environment, constructed in real time by a terminal to accurately understand and respond to a user’s intent when executing a specific task.
3.4 end to end closed loop
The complete process from when a user issues an instruction or intent, through the terminal performing the necessary steps of perception, cognition and execution, to the successful delivery of the task result.
3.5 personal knowledge base
A collection of personalised information about a specific user, such as facts, relationships, preferences and habits, that is learned and stored by a terminal with the user‘s authorisation during long term use.
3.6 session
An interaction process between a user and a terminal to complete one or a set of related tasks.
3.7 session context
A set of dynamic information, recorded and maintained during a session, including the interaction history and current state within that session, to ensure interaction continuity and accurate task execution.
3.8 ability element
A basic category that forms the top level framework of a terminal’s intelligence capabilities.
3.9 task decomposition
The ability of a terminal to automatically break down a complex user intent or a multi step instruction into a series of smaller, more specific and executable sub tasks.
3.10 task orchestration
The ability, based on task decomposition, to plan and generate an optimal or reasonable execution sequence according to the logical relationships, dependencies and execution priorities among sub tasks.
3.11 intent clarification
The ability of a terminal, when it recognises that a user‘s intent is unclear, to further clarify the user’s true need by actively asking questions, providing options or requesting additional information.
4 Abbreviations
The following abbreviation applies to this document.
TTS: Text to Speech
5 Overview
Based on the complexity and degree of automation of the various tasks a terminal can perform within its functional scope, the terminal‘s intelligence level is classified into four levels, namely Level 1 (L1) to Level 4 (L4). The higher the level, the higher the terminal’s intelligence level. Higher level terminals encompass all the capabilities of lower level terminals. For example, an L3 level terminal shall possess all the key capabilities of both L1 and L2 levels.
6 Classification Levels
6.1 L1 Response Level
The terminal can understand a single simple instruction, invoke a specific tool according to the instruction, and complete a single step task.
6.2 L2 Tool Level
The terminal can understand user instructions and simple intents, and has simple reasoning capabilities. It has the ability to invoke pre set tools to complete single step or clearly defined multi step tasks. It has the ability to generate content in at least one modality, such as text, audio or images. It has short term memory within a single session.
6.3 L3 Assistance Level
The terminal can comprehensively understand user instructions and intents, and has the ability to actively clarify intents. It has strong reasoning capabilities and can automatically perform task decomposition and task orchestration. It has the ability to dynamically select and automatically invoke tools, and to generate content in at least one modality, such as text, audio or images. It has both short term and long term memory capabilities.
6.4 L4 Collaboration Level
To be determined.
NOTE: Based on the current state of technological development, this document specifies the key intelligence capabilities for Levels L1 to L3. The classification for Level L4 and above will be defined and refined in subsequent revisions of this document.
7 Capability Elements
7.1 Primary capability elements
7.1.1 Perception
The ability of a terminal to acquire data from sensors, system services and application services through internal/external sensors, data acquisition modules and recognition modules, and to construct the scenario context information required for task execution, including user information perception, device information perception and environmental information perception.
7.1.2 Cognition
The ability of a terminal to understand user intents and related information, to perform reasoning and analysis on information, and to dynamically plan tasks, ensuring that task execution meets the user’s expected goals. This includes understanding, reasoning and planning.
7.1.3 Execution
The ability of a terminal to generate responses based on the output of the cognitive process and to invoke internal/external tools and services to achieve task goals. This includes tool invocation, content generation, interconnection and collaboration, and expression output.
7.1.4 Memory
The ability of a terminal, with the user‘s authorisation, to extract, store, retrieve and dynamically update interaction content and related information. This includes short term memory and long term memory.
7.1.5 Learning
Standard
GB/Z 177.2-2026 Intelligence grading of artificial intelligence terminal—Part 2: General requirements (English Version)
Standard No.
GB/Z 177.2-2026
Status
valid
Language
English
File Format
PDF
Word Count
14500 words
Price(USD)
435.0
Implemented on
2026-4-30
Delivery
via email in 1~5 business day
Detail of GB/Z 177.2-2026
Standard No.
GB/Z 177.2-2026
English Name
Intelligence grading of artificial intelligence terminal—Part 2: General requirements
GB/Z 177.2-2026 Intelligence grading of artificial intelligence terminal—Part 2: General requirements English, Anglais, Englisch, Inglés, えいご
This is a draft translation for reference among interesting stakeholders. The finalized translation (passing through draft translation, self-check, revision and verification) will be delivered upon being ordered.
ICS
CCS
National Standard of the People's Republic of China
GB/Z 177.2-2026
Intelligence grading of artificial intelligence terminal - Part 2: General requirements
人工智能终端智能化分级 第2部分：总体要求
Issue date: 2026-03-31 Implementation date: 2026-10-01
Issued by the General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China
the Standardization Administration of the People's Republic of China
Contents
Foreword
Introduction
1 Scope
2 Normative References
3 Terms and Definitions
4 Abbreviations
5 Overview
6 Classification Levels
6.1 L1 Response Level
6.2 L2 Tool Level
6.3 L3 Assistance Level
6.4 L4 Collaboration Level
7 Capability Elements
7.1 Primary Capability Elements
7.2 Secondary Capability Elements
8 Key Capabilities
8.1 Overview
8.2 End Side Capabilities of High Computing Power Terminals
8.3 Device Cloud Collaboration Capabilities
9 Test Methods
9.1 Test Principles
9.2 Test Framework
9.3 Intelligence Level Determination
9.4 Test Procedure
Annex A (Informative) User Instructions and Intents
Annex B (Informative) Intelligence Level Scoring Model
Bibliography
Artificial intelligence terminal intelligence classification — Part 2: General requirements
1 Scope
This document specifies the classification system for the intelligence level of artificial intelligence terminals (hereinafter referred to as “terminals”), including the terminal intelligence level classification, capability elements and key capabilities for each level, and describes the test methods.
This document applies to the intelligence classification of various types of terminals, and also provides a reference for the design, development, application, selection and testing of terminals.
2 Normative References
The following document is essential for the application of this document. For dated references, only the edition cited applies. For undated references, the latest edition (including any amendments) applies.
GB/Z 177.1 Artificial intelligence terminal intelligence classification — Part 1: Reference framework
3 Terms and Definitions
For the purposes of this document, the terms and definitions given in GB/Z 177.1 and the following apply.
3.1 testing scenario
A comprehensive test context set for testing the key capabilities of a terminal in a specific application scenario.
3.2 testing task
A basic execution unit in terminal intelligence testing.
NOTE: A testing task can serve either as a specific step within a testing scenario or as an independent, atomic test item to be completed separately.
3.3 scenario context
A dynamic set of information about the user, device and environment, constructed in real time by a terminal to accurately understand and respond to a user’s intent when executing a specific task.
3.4 end to end closed loop
The complete process from when a user issues an instruction or intent, through the terminal performing the necessary steps of perception, cognition and execution, to the successful delivery of the task result.
3.5 personal knowledge base
A collection of personalised information about a specific user, such as facts, relationships, preferences and habits, that is learned and stored by a terminal with the user‘s authorisation during long term use.
3.6 session
An interaction process between a user and a terminal to complete one or a set of related tasks.
3.7 session context
A set of dynamic information, recorded and maintained during a session, including the interaction history and current state within that session, to ensure interaction continuity and accurate task execution.
3.8 ability element
A basic category that forms the top level framework of a terminal’s intelligence capabilities.
3.9 task decomposition
The ability of a terminal to automatically break down a complex user intent or a multi step instruction into a series of smaller, more specific and executable sub tasks.
3.10 task orchestration
The ability, based on task decomposition, to plan and generate an optimal or reasonable execution sequence according to the logical relationships, dependencies and execution priorities among sub tasks.
3.11 intent clarification
The ability of a terminal, when it recognises that a user‘s intent is unclear, to further clarify the user’s true need by actively asking questions, providing options or requesting additional information.
4 Abbreviations
The following abbreviation applies to this document.
TTS: Text to Speech
5 Overview
Based on the complexity and degree of automation of the various tasks a terminal can perform within its functional scope, the terminal‘s intelligence level is classified into four levels, namely Level 1 (L1) to Level 4 (L4). The higher the level, the higher the terminal’s intelligence level. Higher level terminals encompass all the capabilities of lower level terminals. For example, an L3 level terminal shall possess all the key capabilities of both L1 and L2 levels.
6 Classification Levels
6.1 L1 Response Level
The terminal can understand a single simple instruction, invoke a specific tool according to the instruction, and complete a single step task.
6.2 L2 Tool Level
The terminal can understand user instructions and simple intents, and has simple reasoning capabilities. It has the ability to invoke pre set tools to complete single step or clearly defined multi step tasks. It has the ability to generate content in at least one modality, such as text, audio or images. It has short term memory within a single session.
6.3 L3 Assistance Level
The terminal can comprehensively understand user instructions and intents, and has the ability to actively clarify intents. It has strong reasoning capabilities and can automatically perform task decomposition and task orchestration. It has the ability to dynamically select and automatically invoke tools, and to generate content in at least one modality, such as text, audio or images. It has both short term and long term memory capabilities.
6.4 L4 Collaboration Level
To be determined.
NOTE: Based on the current state of technological development, this document specifies the key intelligence capabilities for Levels L1 to L3. The classification for Level L4 and above will be defined and refined in subsequent revisions of this document.
7 Capability Elements
7.1 Primary capability elements
7.1.1 Perception
The ability of a terminal to acquire data from sensors, system services and application services through internal/external sensors, data acquisition modules and recognition modules, and to construct the scenario context information required for task execution, including user information perception, device information perception and environmental information perception.
7.1.2 Cognition
The ability of a terminal to understand user intents and related information, to perform reasoning and analysis on information, and to dynamically plan tasks, ensuring that task execution meets the user’s expected goals. This includes understanding, reasoning and planning.
7.1.3 Execution
The ability of a terminal to generate responses based on the output of the cognitive process and to invoke internal/external tools and services to achieve task goals. This includes tool invocation, content generation, interconnection and collaboration, and expression output.
7.1.4 Memory
The ability of a terminal, with the user‘s authorisation, to extract, store, retrieve and dynamically update interaction content and related information. This includes short term memory and long term memory.
7.1.5 Learning
