(
目录 Table of Contents
智能编码行业发展历程与现状 ..... 03 Intelligent coding industry development process and current situation ..... 03
更高效、更普惠，计算机编码发展史 More efficient, more inclusive, the development history of computer coding
"接近自动驾驶"，智能编码的三个阶段 "Approaching Autonomous Driving", Three Stages of Intelligent Coding
CodeGeeX 智能编程助手 ..... 06 CodeGeeX Smart Programming Assistant ..... 06
CodeGeeX 核心能力 CodeGeeX Core Competencies
CodeGeeX 领先优势 CodeGeeX Leading Edge
IDE及编程语言支持 IDE and programming language support
CodeGeeX 部署方案 ..... 17 CodeGeeX deployment solution ..... 17
一体机部署 All-in-one deployment
本地私有化部署 Local private deployment
支持与服务 ..... 18 Support and Services ..... 18
客户成功案例 ..... 19 Customer Success Stories ..... 19
关于智谱AI ..... 22 About Zhipu AI ..... 22
更高效、更普惠，计算机编码发展史 More efficient, more inclusive, the development history of computer coding
"接近自动驾驶"，智能编码的三个阶段 "Approaching Autonomous Driving", Three Stages of Intelligent Coding
CodeGeeX 智能编程助手 ..... 06 CodeGeeX Smart Programming Assistant ..... 06
CodeGeeX 核心能力 CodeGeeX Core Competencies
CodeGeeX 领先优势 CodeGeeX Leading Edge
IDE及编程语言支持 IDE and programming language support
CodeGeeX 部署方案 ..... 17 CodeGeeX deployment solution ..... 17
一体机部署 All-in-one deployment
本地私有化部署 Local private deployment
支持与服务 ..... 18 Support and Services ..... 18
客户成功案例 ..... 19 Customer Success Stories ..... 19
关于智谱AI ..... 22 About Zhipu AI ..... 22
智能编码行业发展历程与现状 Development History and Current Situation of Intelligent Coding Industry
更高效、更普惠，计算机编码发展史 More efficient, more inclusive, the development history of computer coding
从计算机诞生以来，程序设计工作一直在朝着更简单、更低门槛、更符合人类思维方式的方向发展，使得开发效率不断提高，开发成本不断降低：
Since the birth of computers, programming work has been evolving towards simpler, lower thresholds, and more in line with human thinking, resulting in continuously improving development efficiency and decreasing development costs
Since the birth of computers, programming work has been evolving towards simpler, lower thresholds, and more in line with human thinking, resulting in continuously improving development efficiency and decreasing development costs
机器语言和汇编语言：计算机诞生之初，人们需要通过编写机器语言或者汇编语言来直接与计算机硬件进行交互。这些语言非常底层且复杂，需要开发人员具备深厚的硬件知识和计算机原理。
Machine Language and Assembly Language: At the beginning of the birth of computers, people needed to interact directly with computer hardware by writing machine language or assembly language. These languages are very low-level and complex, requiring developers to have a deep understanding of hardware and computer principles.
-面向对象编程：20 世纪80年代，面向对象编程 （OOP）的概念逐渐流行，OOP 通过将数据和操作封装在一起，提供了一种更加直观和易于理解的编程范式，使得开发人员可以将注意力集中在问题本身，而无需过多关注底层实现。
Object-oriented programming: In the 1980s, the concept of object-oriented programming (OOP) gradually became popular. OOP encapsulates data and operations together, providing a more intuitive and easy-to-understand programming paradigm. This allows developers to focus on the problem itself without having to pay too much attention to the underlying implementation.
- 高级编程语言：20 世纪 50 年代起，高级编程语言 （如Fortran、COLBOL和C语言）的出现使得开发人员可以在一个相对抽象的层面上进行编程，这些语言通过提供统一的编程模型和抽象，大大降低了开发难度。
High-level programming languages: Since the 1950s, the emergence of high-level programming languages (such as Fortran, COBOL, and C) has allowed developers to program at a relatively abstract level, greatly reducing development difficulty by providing a unified programming model and abstraction.
低代码和无代码平台：通过提供可视化编程界面和预定义的组件库，使得开发人员无需编写代码，或者只需要编写很少的代码，就可以搭建和实现应用程序。这使得编程变得更加简单、直观和易上手，让更多的人能够参与到软件开发中来。
Low-code and no-code platforms: By providing visual programming interfaces and predefined component libraries, developers can build and implement applications without writing code or with minimal code. This makes programming simpler, more intuitive, and easier to get started with, allowing more people to participate in software development.
Low-code and no-code platforms: By providing visual programming interfaces and predefined component libraries, developers can build and implement applications without writing code or with minimal code. This makes programming simpler, more intuitive, and easier to get started with, allowing more people to participate in software development.
随着人工智能（ AI ）和大模型技术的迅猛发展，及在深度学习和自然语言处理领域取得的重要突破，使得在更广泛的开发场景下由 AI 自动生成或辅助编写代码成为可能，为软件开发带来了新的机遇和挑战。
With the rapid development of artificial intelligence (AI) and large model technology, as well as significant breakthroughs in the fields of deep learning and natural language processing, it has become possible for AI to automatically generate or assist in writing code in a wider range of development scenarios, bringing new opportunities and challenges to software development.
With the rapid development of artificial intelligence (AI) and large model technology, as well as significant breakthroughs in the fields of deep learning and natural language processing, it has become possible for AI to automatically generate or assist in writing code in a wider range of development scenarios, bringing new opportunities and challenges to software development.
从机器语言到人类语言 From machine language to human language
简单化、智能化、降低增效提质 Simplification, intelligence, efficiency improvement and quality enhancement
"接近自动驾驶"，智能编码的三个阶段 "Approaching Autonomous Driving", Three Stages of Intelligent Coding
国际汽车工程师协会针对于智能驾驶制定了六个不同的标准，从 LO 到 L6 分别对应智能驾驶的不同阶段，其中 L0 代表无自动驾驶功能，L6则代表完全自动驾驶，而现阶段的量产车处于接近 L3 的自动驾驶水平。
The International Association of Automotive Engineers has established six different standards for autonomous driving, ranging from LO to L6, corresponding to different stages of autonomous driving. L0 represents no autonomous driving function, while L6 represents fully autonomous driving. Current mass-produced vehicles are at a level close to L3 autonomous driving.
The International Association of Automotive Engineers has established six different standards for autonomous driving, ranging from LO to L6, corresponding to different stages of autonomous driving. L0 represents no autonomous driving function, while L6 represents fully autonomous driving. Current mass-produced vehicles are at a level close to L3 autonomous driving.
类似地，我们可以将智能编码也简单分为从
Similarly, we can also simply divide intelligent encoding into 3 stages from
Similarly, we can also simply divide intelligent encoding into 3 stages from
D LO 阶段代表着所有的代码都需 D LO stage represents that all code needs to be
要由程序员手动编写，可能会借助 Must be written manually by programmers, may be assisted by
IDE提供的提示, 来减少敲击键盘的频次和编程对记忆力的考验。
The prompts provided by the IDE reduce the frequency of keyboard typing and the memory test of programming.
L1 阶段则是程序员可以通过自然语言编写注释或部分代码，作为提示由「神经网络模型」自动填充代码, 从而达到将程序员从编写代码工作中解放出来。
Phase L1 is when programmers can write comments or partial code in natural language as hints for the "neural network model" to automatically fill in the code, thereby freeing programmers from the task of writing code.
L2 阶段则能实现只需要将任务描述清楚，交予AI 模型，模型便能自动地编写相应的程序，并生成相应的功能的效果。
In stage L2, it is possible to achieve the automatic writing of corresponding programs and the generation of corresponding functional effects by simply describing the task and handing it over to the AI model.
要由程序员手动编写，可能会借助 Must be written manually by programmers, may be assisted by
IDE提供的提示, 来减少敲击键盘的频次和编程对记忆力的考验。
The prompts provided by the IDE reduce the frequency of keyboard typing and the memory test of programming.
L1 阶段则是程序员可以通过自然语言编写注释或部分代码，作为提示由「神经网络模型」自动填充代码, 从而达到将程序员从编写代码工作中解放出来。
Phase L1 is when programmers can write comments or partial code in natural language as hints for the "neural network model" to automatically fill in the code, thereby freeing programmers from the task of writing code.
L2 阶段则能实现只需要将任务描述清楚，交予AI 模型，模型便能自动地编写相应的程序，并生成相应的功能的效果。
In stage L2, it is possible to achieve the automatic writing of corresponding programs and the generation of corresponding functional effects by simply describing the task and handing it over to the AI model.
目前 CodeGeeX 及部分国内外主流智能编码软件均处于「L1-L2」之间的发展区间，即可以—定程度上实现代码自动生成，但是也仍需要人工来进行选择与判断。
Currently, CodeGeeX and some mainstream intelligent coding software at home and abroad are in the development range of "L1-L2", which means that code generation can be achieved to a certain extent, but manual selection and judgment are still required.
Currently, CodeGeeX and some mainstream intelligent coding software at home and abroad are in the development range of "L1-L2", which means that code generation can be achieved to a certain extent, but manual selection and judgment are still required.
