CodeGeeX

Short Intro: CodeGeeX is an open-source AI coding assistant developed by Tsinghua University's Knowledge Engineering Group in collaboration with Z.ai, formerly Zhipu AI, one of China's top AI companies with over $1.4 billion in cumulative funding. The model was published at KDD 2023 and introduced HumanEval-X, a multilingual evaluation benchmark the broader research community now uses to assess code generation across languages. The current CodeGeeX4-ALL-9B generation handles code completion, review, unit test generation, repository-level Q&A, and cross-language translation, all at no cost for individual developers.

Key Capabilities:

• Four completion modes covering automatic Stealth, multi-candidate Interactive, cross-language Translation, and template-based Prompt generation

• 100+ programming language support with explicit multilingual training across Python, C++, Java, Go, JavaScript, and dozens more

• Cross-language code translation for converting code between programming languages directly within the IDE

• Repository-level Q&A for answering questions about entire codebases beyond the current open file

• Code review, unit test generation, inline chat, and code interpreter in the CodeGeeX4-ALL-9B model

• Web search and function calling capabilities in the latest model generation

• Cross-file infilling for generating code that spans multiple files within a repository

• IDE integrations for VS Code, JetBrains, and HBuilderX

• Self-hostable open model weights (~26GB) for on-premise deployment on NVIDIA GPU infrastructure

• Private deployment options for organizations requiring complete data control

• Open-source model weights and codebase available through the zai-org GitHub organization