How to reproduce your benchmarks on rented GPUs?

[Mirko185]

I saw your recent GPT-OSS benchmarks on B200 and H100 cards, and I’d like to know if there’s a simple way to run the same benchmark on rented GPUs via Runpod or vastAI.
I'm interested in understanding the potential performance of GPT-OSS 20B (not 120B) on different cards (RTX 5090?) using your engine with a batch size of 32.