We've been working on a new experiment and are excited to share what we've been cooking! Model Fusion is a new feature that analyzes and synthesizes responses from multiple models into one comprehensive answer. Use this for deep research tasks or other interesting use cases you may think of. Let us know what issues you run into, what ideas you have, or any general feedback you may have!
#Give us feedback about Model Fusion!
rigid beacon
tired jungle
When would this be available on the API level?
plucky sluice
I just need to say I really like the phase 2 synthesis as implemented, keep that
elder citrus
In my own local testing, Opus 4.6 nearly always credits GPT 5.4 for better analysis, acknoledging its own oversights.
While GPT 5.4 does not. However it could be due to GPT personalisation setting, where user have a granular range of options.
analog wave
