Description
Ollama is an open-source project for downloading and running large language models locally, maintained under an MIT license on GitHub. As of this writing, the repository has more than 181,000 stars and over 18,000 forks, and is written primarily in Go. Rather than a hosted chat product, Ollama is a runtime: it downloads a model's weights, serves them behind a local API, and lets a person or another program interact with the model directly on their own machine.
Read More
Getting started means installing Ollama on macOS, Windows or Linux (via a one-line install script or manual download), or pulling the official ollama/ollama image from Docker Hub. From there, running "ollama run" with a model name downloads and starts a chat session with a model such as Gemma, Llama, Qwen, DeepSeek, GLM, MiniMax, Mistral or gpt-oss, with the current model catalog listed at ollama.com/library. Ollama also exposes a local REST API for running and managing models programmatically, plus official Python (ollama-python) and JavaScript (ollama-js) libraries so developers can call a running model from their own applications. The project documents its CLI, REST API and a Modelfile format for customizing how models are imported and configured, and it supports llama.cpp as one of its backend inference engines.
Beyond running a single chat session, Ollama can be launched as the local model behind existing coding tools and agents, such as Claude Code, Codex, Copilot CLI, Droid and DeepSeek Harness, via an "ollama launch" command, and the community project OpenClaw uses Ollama to run a personal AI assistant across chat apps like WhatsApp, Telegram, Slack and Discord. A large ecosystem of third-party interfaces builds on top of Ollama's API, including web chat UIs such as Open WebUI, LibreChat, big-AGI and NextChat, and desktop clients such as AnythingLLM, Witsy and Cherry Studio. Ollama has an active Discord community, and its README invites contributors to open a pull request to add their own integrations. Because it runs models on the user's own hardware, the models and speed available depend on local compute and memory rather than a subscription tier.
- CategoryMLOps & Model Deployment
- Pricing ModelOpen source
- DeploymentRuns locally
- PlatformsmacOS, Windows, Linux
Ollama's Core Features
One-command installers for macOS, Windows, Linux and Docker
Local model library covering Llama, Gemma, Qwen, DeepSeek, GLM, MiniMax, Mistral and gpt-oss
"ollama run" CLI for downloading and chatting with a model instantly
Local REST API for running and managing models programmatically
Official Python (ollama-python) and JavaScript (ollama-js) libraries
"ollama launch" integrations with coding agents like Claude Code, Codex and Copilot CLI
Modelfile format for customizing and importing models
Support for llama.cpp as a backend inference engine
Large third-party ecosystem of chat interfaces and desktop clients built on its API
Getting Started with Ollama
Install Ollama: Run the install script or download the app for macOS, Windows or Linux (or pull the Docker image).
Pull a model: Run "ollama run" with a model name, for example gemma4, to download and start chatting with it.
Call it from code: Use the REST API, or the ollama-python or ollama-js library, to send chat requests from your own app.
Launch an integration: Run "ollama launch" with a supported tool name to connect Ollama to a coding agent such as Claude Code or Copilot CLI.
Browse more models: Check ollama.com/library for the full, regularly updated model catalog.
Ollama's Use Cases
- Local LLM chat
- Coding assistant backend
- App integration via API
- Personal AI assistant
