Breno A

Co-founder, Enverge

Builds Enverge Cloud — SSH-native GPU rental on RTX Pro 6000 Blackwell hardware. Writes about comparisons, fine-tuning, and AI infrastructure for teams shipping on Blackwell.

Posts

How many tokens per second can one RTX Pro 6000 serve? — We benchmarked Qwen3.6-35B-A3B FP8 on vLLM on one RTX Pro 6000: 1,493 tok/s at 16 concurrent requests, MTP on vs off, the prefix-cache catch, and cost per request.