vLLM-Omni benchmarks: 2048px image generation in 0.63s on 4x B300
vllm_project · x · 2026-11-10
The vLLM project published inference numbers for vLLM-Omni's image generation and editing (SP4 / text-encoder TP4): on 4× NVIDIA B300 (BF16), 2048×2048 generation takes 0.63s and one-reference editing 0.85s end-to-end; on 4× H200, 0.88s and 1.35s; on 4× RTX PRO 6000 (PCIe, BF16), 1.6s and 2.5s. The benchmarks target the MiniMax-H3-Image model.
Related event: vLLM-Omni and MiniMax-H3-Image speed up image generation(2 posts)→
More from Infra
- Semiconductor selloff looked overdone: order books and hyperscaler spend tell a different story — BenBajarin · 2026-11-11
- KubeRay joins CNCF as the standard way to run Ray workloads on Kubernetes — PyTorch · 2026-11-11
- Hugging Face launches Local AI 301 course on quantization — huggingface · 2026-11-11
- PyTorch megakernel runs 816B MoE on one TPU v7 host at 433 tok/s, beats GB200 on cost — PyTorch · 2026-11-11
- Can local AI kill the Akamai thesis? NPUs ship in a third of new devices, but hybrid wins for now — pdamodaran · 2026-11-11
- AWS official guide: build a Bedrock invocation observability data lake in five stages — AWS ML Blog · 2026-11-11