1m is a good shout! I'm not sure about changing the model, I think Sonnet is generally a good default. But I think it might be nice to make it easier to change the model when you need/want to in a chat.

One option that comes to mind:

• Change the status indicator to a model dropdown when it's in .idle. Default to the settings default
• Also show it when the chat is empty, so you can change the model before sending a first message
• Add model to the chat view model as optional, and use it over the default if set

One downside here would be if you want to send a follow up with a different model. That might be ok - you can stop it if you don't want to wait, then you get idle state and can change the model.

We could also add a model selector in the top right, across from the status bubble. That could either be there all the time (so we have status top left, model selector top right), or we could have it only there when there's a non-idle status taking the left status bubble space, and then come back to the left when idle.

WDYT?

Reworked the PR based on your feedback:

• Kept Sonnet as the default -- agreed it's a better default
• Added per-chat model selector in the status bar -- when idle, the status pill becomes a tappable Menu that lets you switch between Sonnet, Opus, and Haiku for the current chat session. Non-idle states (connecting, streaming, reconnecting, error) still show the normal status indicators.
• 1M context models -- Sonnet and Opus now use the [1m] context variants

The override is ephemeral (per chat session, not persisted). Selecting the same model as the global default clears the override. The picker shows a sparkle icon + model name + chevron to indicate it's tappable.

Looks great, thankyou!