Engine 'e' pushes 10,500 tok/s on a single DGX Station with Nemotron 3.5 Super VL
petrusenko_max · x · 2026-10-17
somewhenesy's new inference engine 'e' is described as "the closest thing to alien technology" they have: it achieves 10,500 tok/s aggregate throughput on a single DGX Station running NVIDIA's just-released Nemotron 3.5 Super VL.
Related event: Inference engine 'e' hits 10,500 tok/s on a single DGX Station(5 posts)→
More from Infra
- Claude Desktop Drains More Battery Than Running Qwen 3.8 Next Locally — andrew_n_carr · 2026-10-17
- Cisco at OCP: the rack is the scale-up edge, Silicon One is the hedge — BenBajarin · 2026-10-17
- Credo's active LED cable extends the AEC playbook into a ~5x larger short-reach optics market — BenBajarin · 2026-10-17
- Broadcom at OCP: Tomahawk 2.0's underrated halves — Ultra and Jericho — BenBajarin · 2026-10-17
- Open Math OS launches as open infrastructure for "Math 2.0" with free compute — LoubnaBenAllal1 · 2026-10-16
- Hyperscalers to pour $690 billion into cloud and AI infrastructure in 2026 — DavidLinthicum · 2026-10-16