MLPerf v6.1 DeepSeek-R1 per-GPU throughput improvement
Company: CoreWeave
The claim, verbatim
CoreWeave increased per-GPU server throughput for DeepSeek-R1-671B by 19.8% from MLPerf v6.0 to v6.1 on NVIDIA GB200 NVL72
Source (primary)
MLPerf® Inference v6.1 Results: CoreWeave Leads Providers - CoreWeave (-, news_article)
View cached copy (2026-09-19)Live source ↗
Quote: “Comparing the 72-GPU v6.1 submission with the 64-GPU v6.0 submission on NVIDIA GB200 NVL72,derived per-GPU Server throughput increased by 19.8%.”
How we checked this
Checked on September 24, 2026. The cited source supports every part of this claim.
CoreWeave's blog states that per-GPU server throughput on GB200 NVL72 rose 19.8% from v6.0 to v6.1, in its DeepSeek-R1-671B section. The post notes that this is a derived figure comparing a 72-GPU submission with a 64-GPU one, and that it is not verified by MLCommons.
Confirmed in the source:
- CoreWeave's per-GPU server throughput for DeepSeek-R1-671B rose by 19.8% between its MLPerf Inference v6.0 and v6.1 submissions
- The comparison was made on NVIDIA GB200 NVL72
What we did: Read our cached copy of the publisher (https://www.coreweave.com/blog/coreweave-leads-cloud-providers-in-mlperf-r-inference-v6-1-performance-with-nvidia-blackwell-ultra) in full (14,189 characters, retrieved September 19, 2026) and checked each assertion in the claim against it.
Additional evidence
confirms MLPerf® Inference v6.1 Results: CoreWeave Leads Providers - CoreWeave
Quote: “Comparing the 72-GPU v6.1 submission with the 64-GPU v6.0 submission on NVIDIA GB200 NVL72,derived per-GPU Server throughput increased by 19.8%.”