How can I make a batch request to query two different models (eg: gpt-3.5-turbo and gpt-4-0613) in the same batch request?