Ollama — локальный запуск языковых моделей

★ 176.6K

Ollama is an open-source tool for running language models on your own machine. Reach for it when the model has to work locally: data must not leave for the cloud, there is no subscription, access is needed offline, or you simply want to experiment for free. It is deliberately simple: one command downloads a model, another starts a conversation — Gemma, LLaMA, DeepSeek, Qwen and dozens more come from a shared catalogue with no hunting for files and parameters. Beyond the command line it exposes a local API compatible with common service formats, so third-party applications and your own code can connect; libraries exist for Python and JavaScript. It runs on macOS, Windows, Linux and in Docker. Speed and quality are bounded by hardware: the more video memory, the larger the model.