Ollama
Run downloaded AI models locally, with a CLI, application and inference API.
Documentation reviewed
Can I use this for my work?
Ollama runs downloaded models through an application, command line and API. Its optional cloud service processes requests remotely and has separate billing.
- Availability
- Available. Runtime available; speech currently Apple Silicon only. Cloud model availability and model-specific terms are separate.
- How to use it
- Desktop app; API. General runtime: macOS, Windows and Linux. Current Phonon packages: Apple Silicon macOS only.
- Processing and data
- On-device option; Self-hosted option; Vendor cloud; Local and cloud options. Local inference and cloud inference have different destinations. Cloud features can be disabled; a local runtime does not make cloud calls local.
- Costs and limits
- MIT runtime; local hardware/model costs; paid cloud use
- Outputs
- Model responses and embeddings where supported; current speech models generate audio. Capabilities depend on the selected model.
The October speech release requires Apple Silicon macOS. Phonon 31M uses Apache 2.0; 40M variants use noncommercial CC BY-NC 4.0. Verify the exact model and hardware.
Compare these facts with another tool →
Publisher documentation was reviewed. ASE has not independently tested this workflow, its outputs, billing, privacy controls or accessibility.
Developers and hands-on users who want a local model runtime and API.
You cannot manage model downloads/hardware, or need current Ollama speech support on Windows or Linux.
About Ollama
Ollama runs downloaded models through a local application, command line and API, with optional cloud-hosted models billed separately. Model capabilities and hardware requirements vary. Current Phonon speech packages require Apple Silicon macOS; the 31M checkpoint uses Apache 2.0 while the 40M variants are noncommercial CC BY-NC 4.0. The MIT runtime license does not replace model licenses or eliminate cloud and hardware costs.