Vercel Connect
Manage runtime credentials and third-party service connections for applications and AI agents across projects and environments.
Ollama
ollama.com
Run open AI models locally or in the cloud, and add model inference or speech generation to applications.
Visit website
Pricing
Freemium
Available on
macOS, Windows, Linux
Category
Developer Tools
What is Ollama?
Ollama runs open AI models on your computer or through its cloud service. Start in the app or terminal, choose a local model, or sign in to access cloud models. It serves developers building AI applications and people who want to run models on their own hardware.
Local execution uses your computer's resources, while cloud models run on Ollama's servers without a local model download. Ollama also supports text-to-speech through Phonon, including streamed playback and audio files. These speech packages require Apple Silicon, despite Ollama's broader operating-system availability.
Key features
Local and cloud execution
Choose models that run on your hardware or use cloud models without downloading their model files.
App and terminal access
Follow setup prompts through the app or terminal. The quickstart also covers connecting desktop applications and coding agents.
Streaming speech
Generate speech from supplied text with Phonon. Play audio as it is produced or redirect terminal output into a WAV file.
Application speech endpoint
Use the OpenAI-compatible /v1/audio/speech endpoint for application integration. Output supports WAV and raw PCM at 24 kHz mono with 16-bit samples.
Local request handling
Local model loading and parallel processing use available memory. Requests can queue when there is insufficient memory to load another model.
Who is it for?
Explore models on a personal computer
A developer can try suitable local models before choosing one for an application prototype.
Add spoken assistant replies
An application developer can use Phonon to read assistant responses aloud on a supported Mac.
Create passage narration
A content creator can turn prepared text into an audio file after checking the selected checkpoint's license.
How to use Ollama
- Download Ollama for your operating system and open the app or launch it from the terminal.
- Follow the setup prompts and choose a local model, or sign in to use cloud models.
- Try the selected model before connecting it to an application.
- For Phonon speech, use Apple Silicon hardware with Ollama 0.40.9 or later and download a suitable checkpoint.
- Supply text for playback, save a WAV file, or connect your application to the speech endpoint.
Ollama pricing
Freemium
Free includes local model execution and starter cloud usage credits. Purchased credits and Pro and Max subscriptions provide additional cloud access. See the official pricing page for current subscriptions and model charges.
Check current pricing ↗What to know before you use it
- Local inference speed depends on hardware. Large models can be slow without a strong GPU, and insufficient memory can cause queuing or overload errors.
- The documented Phonon packages require Apple Silicon and Ollama 0.40.9 or later. They are not covered by Ollama's general Windows and Linux availability.
- Speech API requests accept up to 4,096 input characters. Speech models cannot handle chat, completion or embedding requests, and unsupported speech options return errors.
- Phonon's 31M English checkpoint uses Apache 2.0. The listed 40M checkpoints use CC BY-NC 4.0 and are restricted to noncommercial use.
Frequently asked questions
Does Ollama require cloud inference?
No. You can run models locally. Cloud models are a separate option that runs inference on Ollama's servers.
Which languages does Phonon support?
The default checkpoint is English. Listed 40M packages include English, French, German, Spanish and Portuguese, plus a multilingual variant.
Can I save generated speech?
Yes. Redirect terminal speech output to create a WAV file, or request WAV or raw PCM through the speech API.