Ollama

Run language models on your own machine with one command.

AIOpen source

About Ollama

Ollama runs open weight language models on your own machine, and optionally on Ollama's hosted cloud. Models are pulled down and driven from a command line interface, and the same models can be called over an HTTP API, so one installation serves both a terminal chat and an application.

Installers are published for macOS, Windows and Linux, with the macOS build requiring macOS 14 Sonoma or later, and the whole thing can be installed with a single shell command that pipes the vendor's install script into sh. A model is started with ollama run, and a chat session ends by typing the bye command. The rest of the CLI covers pull, rm, ls, ps, stop, serve, signin, signout and launch. The launch command wires Ollama models into external tools including VS Code and several coding agents. Custom models are defined in a Modelfile and built with ollama create.

The local API listens on port 11434 at localhost. There is also an OpenAI compatible layer that accepts the literal string ollama as the API key and covers chat completions, completions, models, embeddings and responses endpoints. That layer has gaps worth knowing about before you point an existing client at it: it does not support logprobs, logit bias or tool choice, and images must be base64 encoded rather than passed as URLs.

Appending a cloud suffix to a model name runs that model on Ollama's cloud instead of locally, which requires signing in with the ollama signin command or setting an API key in the OLLAMA_API_KEY environment variable. The hosted API is reachable at ollama.com/api, and cloud models are served from the United States, Europe and Singapore. Cloud features can be turned off entirely so that an installation runs only against local models, which then work with no network connection at all.

Cloud use is what the paid plans meter. The Free plan costs nothing, allows one cloud model to run at a time and permits an unlimited number of public models. Pro costs 20 USD per month or 200 USD per year, raises the limit to three cloud models at a time and permits uploading and sharing private models. Max costs 100 USD per month and allows ten cloud models at a time. Team costs 25 USD per seat per month with a five seat minimum and includes zero data retention. Enterprise pricing is quoted rather than listed. Ollama states that user data is not used to train models.

Key features

Runs open weight models on a local machine and, optionally, on Ollama's hosted cloud.
Starts a model with the ollama run command and ends a chat session by typing bye.
Runs a model in the cloud by appending a cloud suffix to the model name.
Turns cloud features off so the installation only uses local models.
Defines a custom model in a Modelfile and builds it with ollama create.
Serves a local HTTP API on port 11434 at localhost, with a hosted API at ollama.com/api.
Exposes an OpenAI compatible API that accepts the literal string ollama as the API key.
Covers chat completions, completions, models, embeddings and responses on the OpenAI compatible layer.
Leaves logprobs, logit bias, tool choice and image URLs out of that compatible layer.
Wires models into VS Code and several coding agents with the ollama launch command.
Provides run, pull, rm, ls, ps, stop, serve, signin, signout and launch as CLI commands.
Installs from a single shell script or from published macOS, Windows and Linux installers.
Serves cloud models from the United States, Europe and Singapore.
States that user data is not used to train models, with zero data retention on the Team plan.

What people use it for

Running a model on a laptop with no network connection at all.
Pointing an existing OpenAI client at a local model by changing the base URL and key.
Packaging a system prompt and parameters into a Modelfile so a team runs an identical setup.
Switching cloud off entirely so prompts never leave the machine.
Calling local models from VS Code through the ollama launch command.
Paying 20 USD a month to run three cloud models at once and share private models.

Key facts about Ollama
Category	AI
Pricing	Open source
Free tier	Yes
Upvotes	12
Listed	22 August 2026

Is this your product?

Claim the listing to correct anything wrong with it, add features and a changelog, and answer questions from people deciding whether to try it. Free, and it stays yours.

Claim this listing

Ollama alternatives

All alternatives

LangChain
The framework most LLM applications are still assembled with.
Open source
Pinecone
Managed vector search, so retrieval is somebody else's operational problem.
Freemium
Replicate
Run open-source models behind an API without touching a GPU.
Paid
Suno
Full songs (vocals, instruments, structure), from a text prompt.
Freemium