[Feature]: Allow model switching in ongoing desktop chat sessions

[kirubel12]

Before submitting

• I searched existing issues and did not find a duplicate.
• I am describing a concrete problem or use case, not just a vague idea.

Area

apps/desktop

Problem or use case

In the current desktop application, users cannot change the AI model in an already ongoing conversation/chat session. When the current model is not performing well, users are forced to either accept poor performance or start a new chat with a different model, losing all context and chat history.

Proposed solution

Implement the ability to switch between available AI models within an ongoing chat session without interrupting or resetting the conversation. This should allow users to seamlessly change models mid-conversation while preserving all existing context and message history.

Why this matters

This enables users to optimize their experience during a conversation by switching to a better-performing model without losing context. It improves flexibility and user productivity, as users can adapt to model performance in real-time without starting over.

Smallest useful scope

Add a model selector or dropdown in the chat interface that allows users to change the active model for subsequent messages while keeping all previous messages and context intact.

Alternatives considered

Currently, users must start a new chat session with a different model, which loses all previous context and conversation history.

Risks or tradeoffs

May require careful handling of system prompts and context windows when switching models with different capabilities. Token usage tracking may need adjustment for multi-model conversations.

Examples or references

No response

Contribution

• I would be open to helping implement this.

[chrissy0]

I'd love this

[orowebregen]

Yes, me too!

[chrissy0]

It's a daily issue for me and a fix doesn't sound too difficult to implement

[ryanhopwood]

We have actually solved this in the product we’re building, you can prompt together with your team and also switch ai without losing context.

[7heMech]

This would be awesome, I wonder what's the blocker, pretty sure opencode has that, though I guess if you use that as provider you can switch models so the issue is switching providers.

[ryanhopwood]

There’s honestly no real blocker, we’ve already implemented it for ours, just finishing the final build features if you want to know more

[juliusmarminge]

Thanks for requesting model switching in ongoing conversations. The orchestrator V2 work has now merged in #2829.

V2 supports changing the provider/model selection in an existing thread. When the change requires a fresh provider session, T3 carries conversation context across through a portable handoff.

Closing this as delivered. If a specific part is still missing in a build containing V2, please open a focused follow-up with the provider/version and the behavior you are seeing.