Batch Process: /v1/chat/completions vs /v1/responses Which Finishes Faster?

I’m currently evaluating batch processing performance across OpenAI endpoints and wanted to get input from others who’ve tested this in real world scenarios

Has anyone compared /v1/chat/completions vs /v1/responses in terms of batch job completion speed?

Would love to hear benchmarks or observations you’ve encountered when choosing between these two