vLLM-Omni supports H3-Image on Vera Rubin
Understand this story
The vLLM-Omni team says it has integrated MiniMax H3-Image into NVIDIA Vera Rubin inference. In its short-prompt serving benchmark, native 4096×4096 text-to-image generation on four GPUs had a median latency of 0.87 seconds; the same request on a single RTX 5090 with CPU layer-by-layer offloading took 52.29 seconds and 61.45 seconds. The team says the H3-Image day-one integration currently uses MiniMax's hosted image API, and it plans to expand support for workflows that deploy H3-Image locally based on vLLM-Omni.
AI overview based on source reports · Updated 2 days ago
Coverage timeline
Follow the coverage to explore different perspectives.
- vLLM BlogFeaturedH3-Image Day-0 Support: 4K in 0.87s with Vera Rubin
vLLM-Omni brought MiniMax H3-Image, an open-weight text-to-image and editing model, to NVIDIA Vera Rubin, reaching 0.87 seconds median for native 4096×4096 generation on four GPUs with SP4 and MXFP8 in a short-prompt serving benchmark. A separate SP1×DP4 throughput configuration delivers 6,115 images/hour, which the authors project to 146,755 images/day and, at an assumed USD 0.06879 per image, USD 105.16/GPU/hour gross revenue at full billable utilization before operating costs. The post also documents VAE decoding gains from CUDA Graph replay plus multiple streams (2.635 s to 0.213 s on four GB300 GPUs) and community validations on AMD, Intel, Huawei Ascend and Moore Threads hardware.
Activity for this story
- Current comparable activity
- 3
- Peak comparable activity
- 8Nov 12 17:00
- Change in the past 24 hours
- -49%
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.