Fluid

A local AI assistant for Apple Silicon Macs that runs Meta Llama 3 offline, with on-device voice input and chat history.

Screenshot of Fluid website

Fluid is a local AI assistant for Mac users who want to ask questions without sending their chats to a hosted model. It runs Meta Llama 3 on your Mac and supports offline conversations.

  • Chat from any app: A global keyboard shortcut opens the assistant while you're working elsewhere on your Mac.
  • Voice input: You can dictate questions, with voice processing kept on your device.
  • Answer controls and history: You can stop an answer, interrupt it with another question, and review earlier messages instead of only the last response.

The closed-source Mac app runs on Apple Silicon with macOS 14 or later. Local inference is available free, uses the Mac's GPU to generate answers, and requires several gigabytes of disk space for the model. The app is idle between requests. Its sandbox prevents it from seeing your screen or accessing your files.

Fluid also offers a separate paid hosted API compatible with part of the OpenAI API specification, including use with Aider. API requests leave your device and run on Fluid's servers in the EU. The service says it doesn't log message content or retain conversations permanently; it temporarily holds them in server memory for processing.

For businesses, FluidBox is a separate server offering that processes company data on premises. It's available to corporate customers in the EU.

Similar to Fluid