Add Options to Finalize Multi-Model Response

The ability to finalize the responses (i.e., synthesize the responses) of multiple models is a very nice feature. However, it appears that the models available to finalize the responses are limited to those that created the responses. It would be nice to choose a different model to finalize the responses. For example, I might initially want to use Gemini Pro 2.5, O3, and Grok 4 to generate the initial responses and DeepSeek R1 to finalize them. I cannot find a way to do this currently.

I would also like the ability to provide my own instructions about how to synthesize the responses. The current “finalize response” is a black box in terms of what the models are being instructed to do. For example, it would be nice if I could assign an agent to finalize the responses and include a system instruction in the agent that reflects my preferences.