Selecting models
Every model from every provider you have connected is available in a single model dropdown at the top of each chat. You can pick the best model for the task at hand, switch mid-conversation, and use different models in different chats — all without leaving the app. To add more models, connect additional providers in Settings → Providers.
The model dropdown
The model selector sits at the top of the chat panel. Click it to open a list of all available models, grouped by provider. Each entry shows the model name and, where applicable, a brief indication of its strengths (speed, context size, reasoning quality).
The list only shows models from providers you have already connected. If you do not see a model you expect, check that the relevant provider is configured in Settings → Providers.
Picking a model when starting a chat
When you open a new chat the model selector defaults to the last model you used. To change it, click the dropdown and choose any model from the list before or after sending your first message. The selected model is shown in the chat header throughout the conversation.
Switching models mid-conversation
You can switch models at any point during an ongoing conversation. Select a different model from the dropdown and send your next message — the app sends the full conversation history to the new model so it has all the context it needs. You do not need to start a new chat.
A few things to keep in mind when switching:
- Context window: different models have different maximum context lengths. If the new model has a smaller context window than the conversation requires, the oldest messages are trimmed automatically to fit.
- Tool support: not every model supports tool use. If you are using tools or agents, choose a model that supports function calling.
- Streaming: responses always stream in real time regardless of which model is selected.
Model tradeoffs
Models vary along several dimensions. The right choice depends on your task:
| Characteristic | Faster / cheaper models | Higher-quality models |
|---|---|---|
| Response speed | Near-instant first token | Slightly slower to start |
| Cost | Lower per-message API cost | Higher per-message API cost |
| Reasoning quality | Good for simple tasks | Better at complex reasoning and nuanced instructions |
| Context window | Varies — often smaller | Varies — flagship models typically offer larger windows |
| Tool use | Supported on most modern models | Generally more reliable function calling |
Match the model to the task
Use a fast, lightweight model (such as GPT-4o mini, Claude 3 Haiku, or Gemini 1.5 Flash) for quick lookups, summarisation, and drafting. Switch to a more capable model (GPT-4o, Claude 3.5 Sonnet, or Gemini 1.5 Pro) when you need deep reasoning, complex code generation, or nuanced analysis. Mixing models across chats is a great way to manage your API usage without sacrificing quality where it matters.Available models depend on your connected providers
The Chat Agent does not bundle any models itself — it uses the API keys you supply. If you want access to a specific model, connect the corresponding provider in Settings → Providers. You can connect as many providers as you like and use different models in different chats at the same time.