ローカルLLM
Unsloth
Run and fine-tune LLMs and diffusion models locally — faster training, smaller memory footprint.
★ 77.2k スターApache-2.0
ガイドを見る →Ollamaは、ollama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: ollama/ollama）。GitHubスターは182,098、フォークは18,090で、MIT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。
The simplest way to run open language models on your own machine.
macOS and Linux use the install script; Windows uses a one-line PowerShell command. Desktop downloads exist too.
Type ollama run followed by a model name — your first run downloads the weights, then you are chatting.
Point coding agents, chat apps, and your own code at Ollama's local API instead of a cloud endpoint.
Pull new models anytime with ollama pull and switch between them freely.
macOS, Windows, Linux, and Docker — up and running in about a minute.
Llama, Qwen, Gemma, DeepSeek, gpt-oss, Kimi, GLM, MiniMax, Mistral, and more, in multiple sizes.
An OpenAI-style local endpoint your apps and agents can call with zero cloud dependency.
Works with Claude Code, Codex, Copilot CLI, OpenCode, OpenClaw, and other harnesses.
Turn it into a personal assistant across WhatsApp, Telegram, Slack, and Discord.
First-party Python and JavaScript libraries plus a big community ecosystem.
macOS / Linux: run the install script. Windows: run the PowerShell one-liner.
curl -fsSL https://ollama.com/install.sh | sh
# Windows PowerShell:
irm https://ollama.com/install.ps1 | iexStart chatting with any model from the library.
ollama run gemma4Call the local API from your app, or launch an integration such as a coding agent.
ollama launch claudeLocal models also power the security shelf: Pentest Harness bills itself as a heaven for hackers — a penetration-testing harness built to run on local models rather than cloud APIs. Also on this shelf: entroly — Cut AI context cost with reversible, byte-exact compression. Readers working through the Local LLM shelf also stop at kotaemon, an open-source RAG-based tool for chatting with your documents.
Yes — MIT licensed, free for personal and commercial use, with no account required.
No — it runs on CPU too, though a GPU makes larger models much faster. Smaller models fly even on modest machines.
Ollama is the easiest way to run models. Unsloth adds training and fine-tuning. Many people use Ollama to run what Unsloth helped them train.
The library lists sizes and strengths per model — start small (a few GB) and scale up if your machine handles it.
Yes — the local API plus official Python/JS libraries make it a popular backend for indie AI apps.
Run and fine-tune LLMs and diffusion models locally — faster training, smaller memory footprint.
A one-click browser for open-source AI apps — discover, install, and run anything.
Alibaba's open voice models: expressive, streaming-capable speech with free-form voice design.
Atomic Agent is a local-first AI agent. Runs open-weight models on your own machine via llama.cpp.
Run AI models locally on your machine with node.js bindings for llama.cpp. Enforce a JSON schema on the model output on the generation level
BitNet — Microsoft's official inference framework for 1-bit large language models, running fast on ordinary CPUs, MIT
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 — Ollama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。
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
LLM inference in C/C++
A high-throughput and memory-efficient inference and serving engine for LLMs
🤯 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Ollama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。
Find the best local models for your hardware — and how to run them.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
Token efficient translation utility built on ChatGPT or any OpenAI compatible LLM.
LLM inference in C/C++
Self-hosted control plane: approve or block rogue agent calls.
A high-throughput and memory-efficient inference and serving engine for LLMs