Ollama lets you run AI models, the kind that power chat assistants, on your own computer. You choose a model, type one short command, and Ollama downloads it and gets it running. Then you can chat with it in Ollama's app or connect it to other apps and tools. Because the model runs on your own machine, your conversations stay private and there's nothing to pay. It works on Mac, Windows and Linux, and Ollama says more than 9 million developers use it.
What it can do
- One command to get started. Tell Ollama which model you want and it downloads it and starts it. It picks a version that's small enough for most computers, so you don't need to understand the technical details.
- A big library of free models. It can run hundreds of openly shared models, including Google's Gemma, Alibaba's Qwen, DeepSeek, OpenAI's gpt-oss, Meta's Llama and Mistral. Some can also look at pictures, and some are built for writing code.
- Private by default. Models you run on your own computer never send your words anywhere. Once a model is downloaded, it works even without an internet connection.
- Plugs into the apps you already use. Ollama runs quietly in the background, and other apps can talk to it. It understands the same requests as OpenAI's service, so many tools built for ChatGPT can switch to Ollama by changing one web address. It also works with Claude Code, Claude Desktop, ChatGPT's Codex and VS Code.
- Sets up AI coding helpers for you. The ollama launch command starts coding assistants such as Claude Code and OpenCode, or personal assistants such as OpenClaw, already connected to the model you pick, with no settings files to edit.
- Borrows a bigger computer when yours isn't enough. The largest models are too big for a laptop. Models marked "cloud" run on Ollama's own servers instead, using exactly the same commands. Ollama says it doesn't use those conversations to train models, and you can switch the cloud off completely.
- Uses your graphics card if you have one. It works with NVIDIA and AMD graphics cards and is especially quick on Apple's M-series Macs. Without a graphics card it still runs, just more slowly.
Where you'll meet it
Ollama is often the engine behind other tools. Open WebUI is the most popular way to give it a chat window that looks and works like ChatGPT. Apps that answer questions about your own documents often pair Ollama, which runs the model, with Chroma, which stores and finds the documents, and LangChain, which joins the pieces together. Under the hood, Ollama is built on llama.cpp.
Ollama, LM Studio or llama.cpp?
These are the three most popular free ways to run AI models on your own computer. They run the same kinds of models at similar speeds, because Ollama and LM Studio both use llama.cpp inside. The difference is how you use them.
OllamaThis pageBest forGetting a model running fast and connecting apps to it
The simplest start: one command downloads and runs a model, and it keeps running in the background for other apps to use. Adds cloud models and one-step setup for coding assistants. You give up a little speed and fine control.
- How you use it
- A simple app, commands, or from other apps
- Open source
- Yes (MIT)
- Cost
- Free on your computer; cloud plans from $20 a month
LM StudioBest forBrowsing and trying models in a friendly app
A polished desktop app with a model shop that shows which models will fit on your computer. Great for trying models side by side without the command line. The app itself is free but not open source.
- How you use it
- A desktop app, with a built-in server for other apps
- Open source
- No
- Cost
- Free
Best forThe most speed and control
The engine the other two are built on. Usually the fastest, runs on almost any hardware and lets you tune every setting, but you choose and download model files yourself and it takes more technical know-how.
- How you use it
- Commands and settings, for technical users
- Open source
- Yes (MIT)
- Cost
- Free
A quick way to choose: want an AI model running on your computer in five minutes, or need one that other apps can use? Pick Ollama. Prefer clicking through a friendly app to compare models? Try LM Studio. Comfortable with technical settings and want every bit of speed? Use llama.cpp. You can't go far wrong, and you can switch later because they all run the same kinds of models.
Good to know
- Your computer sets the limit. Small models run on an ordinary laptop, but bigger, smarter ones need a lot of memory or a good graphics card. If a model is slow, try a smaller version of it.
- Models are big downloads, usually several gigabytes each, so check your free disk space.
- Free models on your own computer aren't as capable as the best paid assistants for hard tasks, but they're good enough for many everyday jobs and improve every few months.
- Ollama's own app is simple. For a fuller chat window with saved chats, document uploads and several users, add Open WebUI.
- Ollama opens its door only to your own computer by default. If you open it up to other machines, put a password or proxy in front of it, because it has no login of its own.
- Cloud models are optional and send your prompts to Ollama's servers. The free plan includes some starter use; running models on your own computer is always free and unlimited.
- Licence: MIT. You can use it for free, including in paid products.