DeepSeek V4.1 Flash hits up to 93.3x per-chip throughput on Vera Rubin NVL72 vs GB300
Xianbao_QIAN · x · 2026-11-17
The vLLM project published the first DeepSeek V4.1 Flash results on NVIDIA's Vera Rubin NVL72, powered by Dynamo + vLLM and measured on SemiAnalysis's AgentX benchmark.
Using InferenceX's performance ruler, compared with GB300 NVL72 running the same Dynamo + vLLM stack:
- 63.4x total token throughput per chip at 400 tok/s/user
- 93.3x at 600 tok/s/user
These are among the first public LLM inference comparisons on the new Rubin platform, highlighting how the vLLM + Dynamo combo scales on next-gen hardware.
Related event: vLLM shares first DeepSeek V4.1 Flash results on NVIDIA Vera Rubin NVL72(5 posts)→
More from Infra
- Ex-Apple Engineer Details Siri Speedups: Cold Starts Cut from 25% to 4% of Requests — jdluk87 · 2026-11-17
- DeepSeek-V4.1-Flash hits 200 tok/s on Vera Rubin as agent ships Engram-table sharing fix — woosuk_k · 2026-11-17
- SpaceX pushes FCC for last 50 MHz used by planes and cars ahead of Nov. 19 vote — RachelVT42 · 2026-11-17
- Hadrian invests $300M to upgrade an 1866 Pennsylvania foundry as its Factory 6 — oyhsu · 2026-11-17
- Pipeline parallelism beats TP2 and DEP2 by 23% throughput for agentic serving on MiniMax-M3.1-Flash — woosuk_k · 2026-11-17
- Harvard study: hyperscale data centers used nearly 2% of US electricity in 2024-2025 — Polymarket · 2026-11-17