# WisdomShell/codeshell

A series of code large language models developed by PKU-KCL

Repository: https://github.com/WisdomShell/codeshell
Canonical: https://ross.abutalabs.com/products/codeshell
Homepage: http://se.pku.edu.cn/kcl
Language: Python
License: NOASSERTION
License Family: other
Last push: 2024-07-18T10:20:27+00:00

## Health v2 (maintenance only)
Score: 27/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 76
- inputs: {"age_days": 1076, "days_push": 776, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1621, forks 116 (observed 2026-08-28T04:05:12.493355+00:00)

## What it is
CodeShell is a family of 7-billion-parameter code large language models developed by PKU-KCL, trained on 500 billion tokens with an 8192-token context window, achieving strong HumanEval and MBPP results among 7B code models. The repository provides model weights (base, chat, and 4-bit quantized chat variants), a C++ runtime for GPU-free local deployment, and VS Code/JetBrains IDE plugins.

## Use cases
- generate code from natural language prompts
- autocomplete code in VS Code or JetBrains IDEs
- run a code LLM locally on a CPU-only 8GB machine
- answer questions about code with a chat model
- quantize a code model to 4-bit for faster inference
- benchmark code models on HumanEval and MBPP
- self-host a coding assistant without sending code to the cloud

## When to choose
- you need a compact 7B code model with competitive benchmark performance
- you want a fully open-stack coding assistant with IDE plugins and local C++ inference
- you must run a code LLM offline or on hardware without a GPU
- you prefer models downloadable from Chinese mirrors like ModelScope or WiseModel

## When to avoid
- you need the absolute best code generation quality regardless of model size
- you require a permissively licensed model for commercial use without reviewing the custom license
- you need a actively updated model with recent training data
- you want a general-purpose assistant rather than a code-focused model

## Facets
- artifact type: library
- maturity: maintenance
- function: llm-inference, machine-learning, code-review, developer-tools
- domain: large-language-models, artificial-intelligence, developer-tools, programming-languages
- platform: python, cpp, cross-platform, cli
- tags: code-llm, code-generation, code-completion, 7b-model, quantization, ide-plugin, vscode, jetbrains, humaneval, mbpp, local-deployment, chat-model, gpu

## Member repositories
- WisdomShell/codeshell (main) score 27

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:12.493355+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:49:05.664974+00:00, confidence not recorded.
  - readme: https://github.com/WisdomShell/codeshell (fetched 2026-08-28T04:05:12.493355+00:00, sha db7bd1a5b905)
  - homepage: http://se.pku.edu.cn/kcl (fetched 2026-08-29T11:21:46.705298+00:00, sha 314501505470)
- Data as of 2026-08-30T08:39:29.467469+00:00.