{"resourceId":"nvidia-vera-rubin-mlperf-v61-20260916","versions":[{"version":"external-c39c2abebe292d4e610c924f51990c8b94a6ad722efaa23c98c01596dd65e714","resource":{"id":"nvidia-vera-rubin-mlperf-v61-20260916","title":"Vera Rubin NVL72 MLPerf Inference v6.1 preview: up to 3.7x vs GB300 on Qwen3-VL","organization":"NVIDIA","sector":"AI infrastructure / inference benchmarking","geography":"Global (vendor MLPerf submission blog)","publishedAt":"September 16, 2026","publicationDate":"2026-09-16","eventDate":"2026-09-16","sourceName":"NVIDIA Blog","sourceLabel":"Vendor MLPerf announcement","sourceUrl":"https://blogs.nvidia.com/blog/vera-rubin-nvl72-mlperf-inference/","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["infrastructure","developers-agents","operating-model"],"finding":"On September 16, 2026 NVIDIA published Vera Rubin NVL72's first MLPerf Inference v6.1 preview results: up to 3.7x higher throughput than GB300 NVL72 on Qwen3-VL (vLLM+Dynamo) across offline/server/interactive, and up to 2.5x on DeepSeek-R1 (TensorRT-LLM). A four-rack GB300 DeepSeek-R1 offline run reached 99% scaling efficiency (72→288 GPUs). Software optimizations delivered up to 1.6x Qwen3-VL gains vs v6.0. The post also cites SemiAnalysis AgentX preview of 30x vs GB300 and notes Nebius as another Vera Rubin preview submitter among 19 partners.","sledRelevance":"Gives SLED buyers a peer-suite preview of next-gen rack throughput and software-stack dependence (Dynamo/vLLM vs TensorRT-LLM), but preview silicon and hyperscale partner lists do not equal campus/agency availability or measured tokens-per-dollar.","evidence":"Blog dated Sep 16, 2026 cites Closed Division entries 6.1-0106/0074 for Vera Rubin vs GB300; 6.1-0073/0074 for four-rack 99% offline scaling; WAN 2.2 0.65 720p videos/s at 5.7s; partner list including Azure, CoreWeave, Dell, HPE, Lambda, OCI, Supermicro; AgentX 30x claim referenced as preview testing outside MLPerf.","architectureImplications":"Expect disaggregated prefill/decode, NVFP4, and full NVL72 NVLink domain to be prerequisites for approaching published multipliers; Jetson Thor Edge-Agentic path is a separate form factor.","governanceImplications":"Treat as preview evidence—require MLCommons entry IDs and scenario matching in procurement evaluations before generational refresh decisions.","securityPrivacyImplications":"No CC or tenancy measurements in this post; do not infer security posture from throughput leadership.","caveats":"Vendor-authored summary of its own submissions; 'up to' spans scenarios/models/stacks; AgentX 30x is not an MLPerf result; no measured system power or rack price.","streamIds":["nvidia"],"roles":{"sales":"Frame Vera Rubin MLPerf as early peer-suite evidence of next-rack inference throughput—not a ready SKU quote. Open with which model and scenario the buyer actually serves, then map to the 3.7x Qwen3-VL (vLLM+Dynamo) versus 2.5x DeepSeek-R1 (TensorRT-LLM) split so stakeholders see stack dependence. Qualify preview status, partner cloud availability, and absence of power/price figures before any refresh conversation.","engineering":"Treat the post as a stack recipe: match framework (vLLM+Dynamo or TensorRT-LLM), precision (NVFP4), and disaggregated serving before estimating uplift. Plan validation on like-sized NVL72 domains rather than mixing single-node and multi-rack numbers. Capture entry IDs 6.1-0106/0074 and scenario tables as the acceptance baseline for any PoC.","delivery":"If a pilot is approved, schedule scenario-matched soak tests (interactive vs server) on current Blackwell first, then compare only when Vera Rubin capacity is contractually available. Document software version pins and disaggregation topology. Do not write acceptance criteria from the AgentX 30x line—keep that outside MLPerf-gated milestones."},"roleDetails":{"sales":{"customerProblem":"Buyers need credible next-gen inference throughput evidence without treating preview blogs as purchase specs.","stakeholders":["CIO/CTO","AI platform owner","procurement","facilities/power"],"discoveryQuestions":["Which model and latency scenario dominate production?","Is the serving stack vLLM/Dynamo or TensorRT-LLM?","Is Vera Rubin capacity actually quotable for your region?"],"valueHypothesis":"Peer-suite preview results can bound expected rack-scale uplift when stack and scenario match.","potentialEngagement":"Scenario-matched readout workshop comparing buyer workload card to published MLPerf entries.","unsupportedClaims":["Guaranteed 3.7x on any model","Lower cost per token proven","Broad commercial availability now"]},"engineering":{"fit":"Fit for teams already on NVL72-class fabrics evaluating Rubin; poor fit for single-GPU or non-NVLink estates expecting the same multipliers.","architectureIntegration":"Reproduce with disaggregated prefill/decode, NVFP4, and full rack NVLink; pin Dynamo/vLLM or TensorRT-LLM versions.","prerequisites":["NVL72 topology","Chosen inference framework expertise","Access to preview/partner capacity"],"constraints":["Preview system","Scenario-specific multipliers","No power submission"],"security":"No attestation or tenancy data in source—keep security evaluation separate.","proofOfValue":["Match within 20% of cited scenario on agreed model","Document framework/precision pins","Separate AgentX claims from MLPerf gates"]},"delivery":{"work":"Baseline current fleet, pin software, run scenario tests, and gate any refresh on availability.","dependencies":["Cloud/OEM capacity","Model owners","Network fabric team"],"ownership":"AI platform engineering lead.","skillsAdoption":"Team needs Dynamo/disaggregation and MLPerf scenario literacy.","governanceCheckpoints":["Preview risk acceptance","Procurement exception for non-GA hardware","Scenario sign-off"],"acceptanceCriteria":["Published entry IDs referenced in test plan","Interactive and server results both captured","No AgentX metric in contractual acceptance"],"risks":["Preview silicon variance","Software stack mismatch","Capacity slip"]}},"retrievedAt":"2026-09-18T03:08:00Z","enrichedAt":"2026-09-18T03:10:00Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Higher inference capacity can improve shared research tooling responsiveness; no accessibility features claimed.","procurementImplications":"Use as evaluation input with mandatory scenario/stack matching; do not sole-source on headline multipliers.","operatingModelImplications":"Encourages treating inference economics as continuous software optimization, not one-time hardware purchase.","sourceVerification":{"openedUrl":"https://blogs.nvidia.com/blog/vera-rubin-nvl72-mlperf-inference/","referenceExcerpt":"NVIDIA Vera Rubin NVL72 delivers up to 3.7x better throughput than GB300 NVL72","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}