Skip to content
AI

Local Models

Pro Exclusive

Local Models let you use AI models running on your own machine through Ollama in Raycast AI. Everything stays local on your own computer. Download a model once and it's yours: it works with no internet connection, keeps every prompt on your machine, and never touches a usage limit. Raycast detects the models you've installed automatically and adds them to the model picker alongside other Raycast AI models.

Requests to local models go straight from Raycast to the Ollama server on your machine. Nothing is sent to Raycast's servers or to any third party.

Get Started

  1. Install the Ollama app if you don't have it yet. You can also click Install Ollama app in Settings → AI → Local Models.
  2. Open Settings → AI and scroll down to the Local Models section.
  3. Type a model name into Install Ollama Model, for example qwen3.8:latest, and click Install. Click Model Library to browse all available models on ollama.com.
  4. Once the download finishes, the model appears in the model picker under the Local section, ready to use.

Models you've downloaded directly with ollama pull are detected too. If one hasn't shown up yet, click Sync Models. The status line at the bottom of the section in Settings shows how many models are installed and when Raycast last synced.

Raycast starts the Ollama server for you when you use a local model, so you don't need to keep the Ollama app open.

Using Local Models

Local models work in Quick AI, AI Chat, and AI Commands. Pick them from the model picker like any other model. Raycast reads each model's capabilities from Ollama, so models that support vision accept image attachments, models that support tool use work with AI Extensions, and thinking models show their reasoning.

Since Raycast has no benchmark data for local models, they don't show speed or intelligence scores in the model settings. Performance depends entirely on your hardware and the model size you choose. Smaller models respond faster, while larger models give better answers at the cost of speed and memory.

Local models don't power the Extension AI API or Emoji Search. Those features continue to use Raycast AI.

Managing Models

  • Install a model from the Local Models section in Settings → AI, or with ollama pull in your terminal.
  • Delete a model from Settings → AI → Models. Click on a local model and choose Delete Model. This removes the download from your disk, so Raycast asks you to confirm first.
  • Disable a model without deleting it by unchecking it in Settings → AI → Models, the same as any other model.

Remote Ollama Servers

If you run Ollama on another machine, such as a home server with a bigger GPU, point Raycast at it with the Ollama Host field in Settings → AI → Local Models. Enter the host in the same form as the default, for example http://192.168.1.20:11434. Leave the field empty to use Ollama on your own machine.

Raycast doesn't start remote servers for you. Make sure the remote Ollama server is running and reachable from your machine, otherwise its models won't load.

FAQ

Troubleshooting

Chat with Raycast AISend Feedback