Ollama

What it does

Ollama downloads open language models such as Gemma, Qwen, DeepSeek or GLM to your computer and runs them there – via the command line, the desktop app or a local REST API. The API is compatible with OpenAI and Anthropic clients, and there are official libraries for Python and JavaScript as well as a Docker image. It supports tool calling, structured outputs, image input and embeddings, for example for your own RAG applications. If your hardware is not powerful enough, you can use larger models through Ollama’s cloud.

Who it suits

Developers and IT teams who want to build or test AI applications without sending data to external services. Ollama also serves as the engine behind interfaces such as Open WebUI or AnythingLLM and for coding agents in the terminal.

Data protection for businesses

When run locally, your inputs stay on your device and Ollama collects no content. For cloud models, prompts and responses are processed only to fulfil the request, not stored and not used for training – there is nothing to switch off. The cloud runs mainly in the United States, with overflow to Europe and Singapore; the vendor does not offer guaranteed EU data hosting. The vendor does not mention a data processing agreement. For companies there is the Team plan and Enterprise with model access controls and security questionnaires.

How we use it

We recommend Ollama in three cases in particular: when sensitive data must not leave the company (such as health or client data), when many requests would get expensive in the cloud, and when an application has to run without an internet connection.

Limits

Which models run locally, and how fast, depends on your hardware; large models are barely usable on basic laptops. Open models do not always match the quality of the best commercial services. Setup requires basic familiarity with the command line or APIs.