Ollama: Run language models locally or in the cloud

Ollama runs supported language models locally or in its cloud, with an app, CLI, APIs, and separate hosted usage billing.

Ollama is an open-source application and CLI for running supported language models on your own hardware. It also offers cloud inference through its app, terminal, and API. Downloaded local models and hosted models have different data paths, hardware requirements, and billing.

Key Highlights

Download and run supported models locally
Connect applications through a local API
Use optional cloud models without downloading their weights
Connect coding agents and compatible API clients
Choose local-only operation by disabling cloud features

What Makes It Different

Ollama supports local inference and hosted inference through related workflows. Local execution can keep model requests on your machine, while cloud requests are processed on hosted infrastructure. The application's open-source license does not make every supported model open source; each model has its own terms.

Features & Capabilities

The app and CLI manage local models and expose APIs for connected tools. Local performance depends on model size, architecture, and your hardware. Cloud models use remote compute and can be accessed through Ollama's API without installing the local application.

Cloud access requires an account or API authentication. Compatible clients support subsets of their original APIs. You can inspect usage and select the model and deployment that suit your workload rather than assuming every request remains offline.

User Ratings and Testimonials

Ollama does not publish a universal comparison score against ChatGPT. Results depend on the selected model, task, and setup. Local use requires suitable hardware; hosted use adds account, network, and usage considerations.

Pricing & Value

The local app and CLI are free. Cloud Free includes starter usage, and purchased credits unlock broader model access. Pro is $20/month or $200/year, Max is $100/month, and Team is $500/month in early access. Paid plans include monthly usage credits, with additional consumption billed as you go. Enterprise pricing is custom. These cloud charges are separate from running downloaded models on your own hardware.