GB/Z 234-2026 人工智能 客服大模型系统技术要求

ICS35.240

CCSL70

中华人民共和国国家标准化指导性技术文件

GB/Z234—2026

人工智能客服大模型系统技术要求

Artificialintelligence—Technicalrequirementsforlarge⁃scalemodelsystemof

customerservice

2026⁃08⁃27发布

国家市场监督管理总局

国家标准化管理委员会发布

GB/Z234—2026

目次

前言··························································································································Ⅲ

1范围·······················································································································1

2规范性引用文件········································································································1

3术语和定义··············································································································1

4缩略语····················································································································1

5框架·······················································································································2

5.1概述·················································································································2

5.2资源层··············································································································2

5.3模型层··············································································································2

5.4平台层··············································································································2

5.5应用层··············································································································3

6模型层····················································································································3

6.1理解能力···········································································································3

6.2推理能力···········································································································4

6.3生成能力···········································································································4

6.4对话能力···········································································································5

7平台层····················································································································5

7.1领域数据···········································································································5

7.2应用支撑···········································································································6

7.3能力扩展···········································································································6

8应用层····················································································································7

8.1客户交互···········································································································7

8.2坐席交互···········································································································8

8.3运营支撑··········································································································10

参考文献····················································································································12

Ⅰ

GB/Z234—2026

前言

本文件为规范类指导性技术文件。

本文件按照GB/T1.1—2020《标准化工作导则第1部分：标准化文件的结构和起草规则》的规定

起草。

请注意本文件的某些内容可能涉及专利。本文件的发布机构不承担识别专利的责任。

本文件由全国信息技术标准化技术委员会（SAC/TC28）提出并归口。

本文件起草单位：中国电子技术标准化研究院、中国移动通信集团有限公司、中国联合网络通信集

团有限公司、中移在线服务有限公司、南方电网人工智能科技有限公司、联通数据智能有限公司、中国

科学院声学研究所、广东电网有限责任公司供电服务中心、中国民航信息网络股份有限公司、中电信数

智科技有限公司、中移九天人工智能科技（北京）有限公司、美的集团（上海）有限公司、中国电信股份有

限公司北京研究院、中移互联网有限公司、浪潮电子信息产业股份有限公司、咪咕文化科技有限公司、

中国电子技术标准化研究院华东分院、浙江大华技术股份有限公司、浪潮通信信息系统有限公司、北京

浩瀚深度信息技术股份有限公司。

本文件主要起草人：冯俊兰、宋文林、徐洋、董建、秦日臻、杜宇、刘阳、马俊鹏、杨森轲、吴石松、

李轩昂、廉士国、王恺、赵庆卫、周军、曾晓锋、杨英勃、于明亮、李劲松、陈利明、邢驰、邓超、孟繁宇、

高羽、刘晴、郑正广、邸贺亮、李琳、马丽萌、董月文、孔维生、肖红梅、李泉、庞韶敏。

Ⅲ

GB/Z234—2026

人工智能客服大模型系统技术要求

1范围

本文件确立了客服大模型系统的框架，规定了模型层、平台层、应用层的技术要求。

本文件适用于客服大模型系统的设计、开发、测试、应用和运维过程。

2规范性引用文件

下列文件中的内容通过文中的规范性引用而构成本文件必不可少的条款。其中，注日期的引用文

件，仅该日期对应的版本适用于本文件；不注日期的引用文件，其最新版本（包括所有的修改单）适用于

本文件。

GB/T41867信息技术人工智能术语

GB/T45288.2—2025人工智能大模型第2部分：评测指标与方法

3术语和定义

GB/T41867界定的以及下列术语和定义适用于本文件。

3.1

大模型large⁃scalemodel

大规模深度学习模型large⁃scaledeeplearningmodel

基于大量数据训练得到，具有复杂计算架构，能处理复杂任务，且具备一定泛化性的深度学习

模型。

注：大模型的参数量由其功能和模态决定，一般不低于1亿。大模型训练使用的数据总量受参数量的影响，达到收

敛的大模型的参数量的对数与其训练数据总量的对数成正比。

[来源：GB/T45288.1—2025，3.1]

3.2

客服大模型large⁃scalemodelofcustomerservice

面向智能客服业务的专业大模型，相比于通用大模型有更高的智能客服业务场景适配度。

3.3

客服大模型系统large⁃scalemodelsystemofcustomerservice

由客服大模型、专业领域知识、应用支撑组件、能力扩展组件等构成的大模型系统。

4缩略语

下列缩略语适用于本文件。

API：应用编程接口（ApplicationProgrammingInterface）

APP：应用软件（APPlication）

OCR：光学字符识别（OpticalCharacterRecognition）

1