Hi all! We just released a new feature for in-conversation fast model choice! This makes it super-easy to switch between models, you can take a look here: https://www.openhands.dev/blog/model-choice-llm-profiles

Thanks for sharing this - it is a great feature to have! One question I had was about the frequency at which a stronger model would by itself offload the task to a weaker model (atleast a few generations back, frontier models had trust issues and would rarely do this well if prompted afaik)? Another question/idea I have is that can we now ask CodeScout to do localization and Claude/GPT to edit more easily than before (I think this was possible previously as well by simply calling conversation.run() twice and switching the model mid-way?)?

Yes, it should be very easy to do a demo video of how we can use codescout with this functionality! You could try downloading the agent-canvas, setting up a codescout profile, and trying i tout