Can I use multiple models for the same thread?

API
,
1

Hello,

I created 2 assistant IDs with the only difference being the model: one uses GPT-4o and the other GPT-4o-mini.
A user can start a conversation with assistant ID 1 and create a thread for it. However, I want to give the user the option to switch models mid-conversation, for example, from GPT-4o-mini to GPT-4o.

Ex :

Assistant ID 1 : Gpt4o-mini
Assistant ID 2 : Gpt4o

  • A user asked a question using the GPT-4o-mini model (I used assistant ID 1 to create a thread in the background).

  • During the same conversation, the user clicked a button to change the model from GPT-4o-mini to GPT-4o.

Should I switch the assistant ID from 1 to 2 and create a new thread for this model change, or can I continue using assistant ID 1 and the existing thread Even though the model is different?

2

whatever assistant that is run on the thread is the model that is used.

3

Please i don’t understand your answer very well. Inside the thread I cannot determine the model.

4

But you can add another assistant if you want, when you will create a new run with a new assistant.id + his instructions, model, other settings.

But is expensive, you should use multiple threads to make some economy.

5

Maybe im not understanding your setup. I would just add both agents to the same thead and you can talk to either one, and they can talk to eachother if u want.

7

threads are assistant agnostic. you do not need assistant id to create thread. this means you can use any assistant during each run. however, whether this is good or bad in your case is another thing.

2 Likes
8

Wow, pretty cool, nice work. What is the front end on that interface? You should check out vrsen/agency-swarm on github, it is a very well designed multi-agent framework capable to be used in production.