Sign in Submit site

CodeGeeX

github.com/zai-org/CodeGeeX Models & ML Research Open-source projectOpen source

Use CodeGeeX, an open multilingual code generation model described in its research project.

Visit CodeGeeX 24 alternatives
github.com/zai-org/CodeGeeX
CodeGeeX homepage screenshot

About CodeGeeX

CodeGeeX is an open-source code generation model with 13 billion parameters, described in its research project as multilingual. The source says it was pre-trained on a code corpus covering more than 20 programming languages. Its main uses are generating code in several mainstream languages, translating code snippets between languages, and running a programming assistant in the VS Code extension marketplace. The repository also publishes the HumanEval-X benchmark for evaluating multilingual code generation and translation.

The code and model weights are publicly available for research purposes, according to the source. Weights are obtained through an application step, and the source lists the hardware it supports for inference (single Ascend 910, NVIDIA V100, or A100). The homepage also refers to newer releases, CodeGeeX2 and CodeGeeX4, so readers should check the repository for the current version.

Key features

  • Generates code in Python, C++, Java, JavaScript, Go, and other languages
  • Translates code snippets between different programming languages
  • VS Code extension offering code completion, explanation, and summarization
  • Supports Ascend and NVIDIA hardware platforms for inference
  • HumanEval-X benchmark with 820 coding problems in five languages
  • Quantization and model parallelism options to reduce GPU memory requirements
Best for: Researchers and developers working on multilingual code generation or code translation who have access to the supported GPU or Ascend hardware.
Good for
  • Code and model weights are publicly available
  • The VS Code extension is listed as free in the marketplace
  • The HumanEval-X benchmark is released for standardized multilingual evaluation
Watch out for
  • The source states that code and model weights are available for research purposes
  • Model weights require an application step before use
  • Inference is listed for specific high-end hardware, so local use depends on available GPUs or Ascend processors

CodeGeeX at a glance

Type
Open-source project
Pricing
Not confirmed. Check the site
Open source
Yes · Source code
Language
English
Last checked
Oct 6, 2026 · Online when checked
Listed
Oct 5, 2026
open sourceml researchcode generationmultilingual

CodeGeeX alternatives

Tools and sites you could use instead, ranked by how closely they match.

All 24 alternatives

Popular in Models & ML Research

Well-known listings from the same topic.

All 69
Falcon LLMfalconllm.tii.aeERNIEernie.baidu.comTokenVersetoken-verse.github.iowTAObittensor.comAlpacacrfm.stanford.edu/2023/03/13/alpaca.htmlSmolVLAhuggingface.co/blog/smolvla
Related topicsAI Business ToolsAI Image GeneratorsAI Infrastructure & MLOpsAI Agents & AutomationAI Data & AnalyticsAI Chatbots & AssistantsAI Voice & AudioAI Writing Tools

CodeGeeX FAQ

Which programming languages does HumanEval-X cover?
HumanEval-X contains 820 human-crafted coding problems in five programming languages: Python, C++, Java, JavaScript, and Go. Each problem is associated with tests and solutions.
What hardware does the source list for running inference?
The source lists inference on a single Ascend 910, NVIDIA V100, or NVIDIA A100. It also states that the code supports both Ascend and NVIDIA platforms.
How can the model's GPU memory requirements be reduced?
The news section says a quantization version requires less GPU RAM, reduced from 27GB to 15GB, and that model parallelism can run the model across multiple GPUs with less than 8GB of RAM each.
Is something wrong or out of date?

Tell us and an editor will check it.

Own this website?

Claim the listing and suggest updates yourself. An editor checks every claim first.

Claim this listing

Listing maintained by Chris McClure and the WebIndexly editors · Updated Oct 13, 2026 · How we rank