Radar check
What we found when we opened it. Every bar is one fact, with its date.
Last code pushed 28 Oct 2026, the same day before this check.
Latest release v1.123.1 on 13 Oct 2026.
Public since March 2023, a little over 3 years.
12,083 stars and 802 forks.
1,748 commits and 104 new issues in the last 90 days.
2,180 contributors. 51 pull requests merged in 90 days.
No security policy file says how to report a problem. 2 published security advisories.
How the bars are scored
- Alive. Days since the last code push: a week or less 100, a month 90, three months 70, six months 50, a year 30, eighteen months 15, older 0.
- Releases. Age of the latest GitHub release, on a similar scale; no bar when the project publishes no releases.
- Maturity. Age of the project: eight years 100, five 90, three 70, two 55, one 40, six months 25, younger 10.
- Adoption. Stars on a log scale: 100,000 stars is 100, 10,000 is 80, 1,000 is 60, 100 is 40.
- Activity. Commits in the last 90 days: 300 or more 100, 100 85, 30 65, 10 45, any 25.
- Community. People who have contributed: 100 or more 100, 30 80, 10 60, 5 40, 2 25, one 10.
- Security. A published security policy 80, none 40; published advisories are listed, not scored.
Checked by Repo Radar on 28 Oct 2026 from the repository on GitHub.
In simple English
This tool lets you run powerful artificial intelligence models right on your own computer. You download one file, pick a model file, and start a local web page in your browser. It chats with you, writes stories, and generates images or voice clips without sending your data to the cloud. You do not need to install complex programming tools to use it. It is free to run, though you need a computer with enough memory to hold the model.
What it is
KoboldCpp is a single-file program that packages a chat interface and an AI engine into one executable. It runs text, image, and audio models on your computer's CPU or graphics card.
What it is good for
Solo builders who want to run private AI models locally for chatbots, writing tools, or local API development without paying monthly subscription fees.
Summary (Key Points)from the README
Key features
- Local text generation: Runs any GGUF language model in chat, adventure, instruct, or story modes, and compatible vision models add image understanding.
- Image and video generation: Supports Stable Diffusion 1.5, SDXL, SD3, and Flux for images, plus WAN 2.2 and LTX2.3 for video.
- Speech and music: Whisper handles speech-to-text, Kokoro and Qwen3TTS handle text-to-speech, and ACE Step 1.5 handles music generation.
- Tools and agents: Supports MCP servers, tool calling, web search, and RAG through TextDB, plus an integrated KoboldCpp Agent that starts with `--agent`.
- Writing and roleplay: The bundled KoboldAI Lite WebUI offers memory, world info, author's notes, and Tavern character card import.
- Compatible APIs: Endpoints for KoboldAI, OpenAI, Ollama, A1111/Forge, ComfyUI, Whisper, and XTTS are served at the default `http://localhost:5001`.
- Portable executable: One file for Windows, Linux, or macOS needs no installation and runs on CPU or GPU with full or partial offloading.
- Hosted provider mode: Launching with `--nomodel` connects the web UI to AI Horde, OpenAI Compatible, Anthropic, OpenRouter, Gemini, Grok, or Mistral instead of a local model.
Ways to install
| Platform or method | What it needs | Effort |
|---|---|---|
| Windows executable | Download `koboldcpp.exe` and double-click it | Low |
| Linux x86-64 binary | `chmod +x`, then run from a terminal | Low |
| macOS (Apple Silicon) | `chmod +x`, and possibly a security whitelist step | Low |
| Android via Termux | Termux from F-Droid, then the installer script or a manual build with a small GGUF model | Medium |
| Source build | Git, Python 3, a C/C++ toolchain, and GPU backend libraries | High |
| Google Colab | A Google account, and compliance with Colab's terms | Low |
| Docker | Experts only, with an NVIDIA or AMD GPU expected | High |
Important
- Models not included: KoboldCpp ships without model files, so you must download a GGUF model separately.
- Phishing risk: `koboldcpp.com` is a malicious fake site. Only trust the official GitHub releases or koboldcpp.net.
- Hardware limits: Memory needs depend on model size and context length, and autofit is only an estimate that may need manual tuning.
- Old CPUs: Some devices crash at startup; try the `oldpc` build or `--noavx2`.
- Intel Macs: These require a source build.
- Docker: It may perform poorly on some Windows or macOS setups, and ARM systems may fail to run it.
- Android: The instructions cover CPU builds only, and GPU acceleration depends on the device and its drivers.
- Third-party projects: Unofficial packages and forks may be outdated or unmaintained.
- Licence: KoboldCpp and KoboldAI Lite use AGPL v3.0, while bundled components keep their own licences.
Written by an AI model from the project’s README; not part of the review, and it can be wrong.
- Runs entirely offline without sending your private data to external servers.
- Packaged as a single standalone executable file with zero complex installation steps.
- Provides compatible API endpoints so other apps can talk to your local model as if it were OpenAI.
- Requires a computer with a dedicated graphics card and plenty of RAM to run large models smoothly.
- Downloading and selecting the right model files and hardware settings can be confusing for beginners.
- Uses the AGPL-3.0 licence (a strict open-source licence that requires you to share your source code if you modify the software and host it publicly).
Paid cloud AI chat subscriptions and complex local Python AI environments
A GGUF model file downloaded from Hugging Face
- ollama/ollama185k stars
Get up and running with large language models locally via a simpler command-line tool.
- ggml-org/llama.cpp133k stars
The low-level C++ inference engine that powers the model execution underneath KoboldCpp.
- vllm-project/vllm96k stars
A serving engine that makes one graphics card answer many requests at once.
- infiniflow/ragflow93k stars
A question-answering system built on a document collection, with the sources shown beside every answer.
- basicmachines-co/basic-memory4.3k stars
Gives an AI assistant a memory that is stored as plain text files you can read.
curl -fLo koboldcpp-linux-x64 https://github.com/LostRuins/koboldcpp/releases/latest/download/koboldcpp-linux-x64 chmod +x koboldcpp-linux-x64 ./koboldcpp-linux-x64