ComputeData release
NVIDIA and Microsoft Azure run the DeepSeek-V3 MLPerf benchmark at the round's largest scale of 10,240 GPUs
NVIDIA used GB300 systems and Azure used GB200 systems at 10,240 GPUs. Vera Rubin NVL72 scaled from 256 to 1,024 GPUs with about 97% efficiency, cutting training time from 16 minutes to about four.
- Size
- 10,240 GPUs
- Companies
- Microsoft Azure, NVIDIA
- Source
- Nvidia news post, published Nov 16, 2026
- Published here
This fact was extracted from the source above and checked automatically: every number and company named here appears in it. When the same fact turns up in another filing or post, that source is added here instead of a second copy. Read the original before relying on it.