vLLM benchmarks MiniMax-M3.1-Flash on Vera Rubin NVL72: 10.2x per-chip throughput vs GB300
vllm_project · x · 2026-11-12
The vLLM team reports continued optimization of MiniMax-M3.1-Flash on NVIDIA Vera Rubin NVL72, using SemiAnalysis AgentX data (KV offload on):
- 10.2x total tok/s/chip (including cached tokens) vs GB300 NVL72 at 546 P90 tok/s/user
- 6.5x at 376 P90 tok/s/user
- Up to 1,205 P90 tok/s/user at low concurrency
The throughput-vs-interactivity curves from SemiAnalysis InferenceX show Rubin (Dynamo vLLM) clearly ahead of GB300 across the range — among the first public inference comparisons of the next-gen Rubin platform against its predecessor.
Related event: vLLM Benchmarks Show Rubin NVL72 Delivering 10.2x GB300 Throughput(2 posts)→
More from Infra
- 768 tiny Pollen robots walk a simulated hospital ward on one GPU — MaziyarPanahi · 2026-11-12
- celld self-hosts Workers, Durable Objects and queues using just an object store bucket — irvinebroque · 2026-11-12
- Azar: transparent chip diffusion strategy backfired as China pivots models to Huawei chips — beffjezos · 2026-11-12
- Running a minimal local agent on 24GB RAM: Pi Agent + Bonsai 2 9B hits ~30 tok/s — Potential_Block4598 · 2026-11-12
- DigitalOcean: Many LLM Serving "Wins" Are Noise — One +4% Winner Wasn't Real — AI Engineer · 2026-11-12
- Homemade inference box pairs AMD CPUs with Huawei Ascend 950PR NPUs — AIFlow_ML · 2026-11-12