Whether you want the best all-around assistant, a model you're already familiar with, or one that fits a specific type of task, you can switch which model Coworker uses directly from the Work page.
How to switch models
Go to the Work page.
Find the model dropdown in the chat interface (top or bottom of the input area).
Click it to see the available models.
Select the model you want — the current session will use that model going forward.
Note: Your model selection applies to the current session. It resets to the default (Claude Sonnet) after a page refresh.
Available models
Three models are currently available:
Claude Sonnet (default) — Anthropic's flagship model. Balanced across reasoning, writing, and analysis. This is the recommended starting point for most tasks.
ChatGPT — OpenAI's model. Strong at conversational tasks, structured outputs, and general-purpose writing.
Gemini — Google's model. Well-suited for tasks involving large amounts of text or when you want a different perspective on a problem.
Which model should I use?
All three models work with your organizational data via Coworker's OM1 layer, so the core experience is consistent regardless of which you choose. That said, each model has different strengths:
Task type | Recommended model |
General research and analysis | Claude Sonnet (default) |
Writing, summarization, and editing | Claude Sonnet or ChatGPT |
Structured data outputs (tables, lists) | ChatGPT or Claude Sonnet |
Processing large bodies of text | Gemini |
You prefer a specific model from experience | Use whichever you know best |
Comparing outputs across models | Try the same prompt in 2–3 models |
When in doubt, start with Claude Sonnet — it's the default for a reason and handles the widest range of tasks reliably.
Tips
All models work in both Auto and Deep Work modes. You can switch the model regardless of which chat mode you're using.
Your selection persists within a session. If you switch to ChatGPT mid-conversation, Coworker will continue using it for subsequent messages until you change it again or refresh the page.
Historical chats always display Claude Sonnet as the model — this is a known limitation and does not reflect which model was actually used.
If one model feels slow or unresponsive, try switching to another. Coworker includes fallback logic to maintain reliability, but switching manually is always an option.