Ollama

Ollama is an API and app for running open models locally or in the cloud, with a stated promise that prompts are never stored or trained on. It targets developers who want private, flexible inference.

Ollama is a platform for working with open models, offered as an API that can run inference on your own machine or in the cloud. The homepage positions it as the most popular way to build with open models, and backs that with usage figures: more than 9 million installs a month, over 1 billion model downloads and more than 200 trillion tokens served. The central promise is privacy. According to the site, prompts are never stored or trained on, whether the model runs locally or through the hosted option. That makes it relevant for developers, researchers and teams who handle sensitive text and want to avoid sending it to a closed provider. Getting started is split into two entry points: a signup flow for the cloud side and a download for the local application. In practice, the tool acts as the layer between an application and an open model. A developer installs it, pulls a model from the catalog of open releases, and sends requests to it the same way they would call any hosted inference API. Because the same interface covers local and cloud inference, a project can start on a laptop or workstation and later move heavier workloads to hosted hardware without rewriting the integration. The homepage describes this as inference anywhere you already work, though it does not list specific editors, frameworks or operating systems. Among alternatives, Ollama sits closer to infrastructure than to a finished chat product. Closed model APIs offer convenience and often stronger frontier models, while Ollama trades that for control over where data goes and which open model is used. The homepage lists well-known organizations in a logo strip, including Apple, NVIDIA, Microsoft, Meta, Adobe, Intel, NASA, Netflix, IBM, Nike, Visa and Mercedes-Benz, but it does not explain how each one uses the product. Pricing, supported platforms and model lists are not stated on the homepage, so buyers need to check those details before committing to a deployment.