ローカルLLM
llama.cpp
LLM inference in C/C++
★ 130.2k スターMIT
ガイドを見る →vLLMは、vllm-projectが公開しているオープンソースのローカルLLMプロジェクトです（GitHub: vllm-project/vllm）。GitHubスターは93,120、フォークは22,958で、Apache-2.0ライセンスの下で公開されています。GitHubでの最終更新は2026年10月3日です。
A high-throughput and memory-efficient inference and serving engine for LLMs
One-command installers exist for every major OS; servers and engines use Docker or simple binaries.
Pull from a built-in library or point at any GGUF/compatible weights you already have.
Use the bundled chat UI, or point your apps and coding agents at the local OpenAI-style endpoint.
Pick model sizes that fit your RAM/VRAM; quantization options trade a little quality for a lot of speed.
Prompts never leave your machine — no accounts, no logging, no training on your data.
Once weights are downloaded, everything runs with the network cable unplugged.
Local endpoints your existing tools and code can call with minimal changes.
Llama, Qwen, Gemma, DeepSeek, Mistral and more, in sizes from phone-friendly to workstation-class.
Coding assistants and automation tools plug straight into the local server.
Run bigger models on smaller hardware via GGUF and friends.
Download the source and move into the project folder.
git clone https://github.com/vllm-project/vllm.git
cd vllmIsolate dependencies so nothing clashes with your other projects.
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activateInstall the requirements, then follow the README's launch command (it varies per project).
pip install -r requirements.txtAlso on this shelf: ai-file-sorter — AI file organizer that sorts and renames files by understanding content. The Local LLM shelf also holds OpenOSINT — AI OSINT agent with 20 tools and REPL — for authorized research only. Readers working through the Local LLM shelf also stop at subsai, Whisper-powered subtitle generation: Web UI, CLI, and package.
Roughly the model size in GB plus headroom — 8B models fit comfortably on 16GB machines with quantization.
Local wins on privacy and cost; cloud wins on raw frontier quality. Many people use both.
It is an open-source project at vllm-project/vllm with 22,958 forks. Check the repo's contributors and recent commits to see how active maintenance is.
A recent mid-size open model in the library's recommendations — start small and scale up.
Yes — the local API plus client libraries make it a popular backend for indie apps.
Yes — several of these are optimized for it, and unified memory helps a lot.
LLM inference in C/C++
High-performance In-browser LLM Inference Engine
Open-source desktop app for local LLMs. Text, vision, tool-calling, OpenAI/Anthropic-compatible API. 100% private
Alibaba's lightweight, blazing-fast inference engine for on-device LLMs and edge AI.
vLLM-Omni — efficient inference for omni-modality models: image, audio, video, diffusion and policies.
TokenSpeed — a speed-of-light LLM inference engine designed for agentic workloads, with TensorRT-LLM-level performance and vLLM-level usability
LibreAIカタログに掲載されたローカルLLMプロジェクトの中で最もスターの多いもの — vLLMが合わなかった場合に検討したい、現実的な候補リストです。
The simplest way to run open language models on your own machine.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
LLM inference in C/C++
🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team.
GPT4All — run local LLMs on any device: a free desktop chat client and Python SDK that run open models entirely on your CPU, no GPU or cloud required
カタログの中でvLLMとスター数が近いローカルLLMプロジェクト — 前後の棚もそのまま見て回れます。
LLM inference in C/C++
🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
GPT4All — run local LLMs on any device: a free desktop chat client and Python SDK that run open models entirely on your CPU, no GPU or cloud required
The simplest way to run open language models on your own machine.
Run and fine-tune LLMs and diffusion models locally — faster training, smaller memory footprint.