Ollama

Ollama runs open models on your computer. Duet Translate can use it like any other provider: free, private and offline once the model is downloaded.

1. Install Ollama and a model

1. Download Ollama from ollama.com and start it.
2. Download a model in a terminal, for example:

ollama pull qwen3:8b

Pick a model your computer can run comfortably. Models with 7 to 14 billion parameters translate well on a recent laptop; larger ones need a strong GPU.

2. Let the browser reach it

Ollama only answers requests from origins it trusts. If the connection test fails with an error such as 403, allow browser extensions and restart Ollama:

macOS

launchctl setenv OLLAMA_ORIGINS "chrome-extension://*"

Then quit Ollama from the menu bar and open it again.

Windows: add a user environment variable named OLLAMA_ORIGINS with the value chrome-extension://*, then quit Ollama from the taskbar and start it again.

Linux (systemd): run sudo systemctl edit ollama.service, add the lines below, then sudo systemctl restart ollama.

[Service]
Environment="OLLAMA_ORIGINS=chrome-extension://*"

3. Add it in Duet Translate

1. On API Providers, click Add provider and choose Ollama.
2. Leave the base URL as http://127.0.0.1:11434/ unless Ollama runs on another computer.
3. Click Fetch available models and pick the one you downloaded.
4. Click Test connection, then choose the features that should use it.

Tips

• Local models are slower than cloud services. Lower the Maximum paragraphs per request under Translation queue if translations time out.
• Thinking is turned off for Ollama models, so they answer right away.