MLCommons

MLPerf Inference v6.0 is here - our most significant benchmark update ever. 5 new/updated benchmarks. 24 submitting organizations. Industry-first tests for text-to-video and speculative decoding. Full results: mlcommons.org/2026/04/mlperf… #MLPerf #MLCommons

Super excited to share that MLPerf Power (HPCA 2025) was selected for IEEE MICRO Top Picks 2025, 1 of the 12 most impactful computer architecture & systems papers of the year! Power consumption is the defining constraint for modern ML systems. Microsoft, Google, Amazon, Meta, and OpenAI have all announced plans for gigawatt-scale datacenters (for context, 5 GW = 5 nuclear reactors = Miami's power footprint). On the other end of the spectrum, we're anticipating billions of AI-enabled devices at the edge. We created MLPerf Power to be the industry-standard to measure, understand, and compare energy use across all deployment scales. We're excited to see that it's already impacting individual companies' strategies and has been incorporated into the IEEE semiconductor roadmap. We @MLCommons also collect and open source over 1,800 reproducible measurements from 60 diverse systems. These reveal several important insights that shed light on the nonlinear scaling of energy efficiency in modern systems and can enable many new data-driven optimization approaches. Just as @MLPerf aligned industry towards shared performance goals, we are hopeful that MLPerf Power will do the same for power and energy efficiency!

@hanlintang Tejus, myself, Sachin, and a few others started MLPerf Power a long time ago and it's amazing to see what it has blossomed into!!! Kudos to everyone on all the work and it's fantastic to see the insights we are now yielding.

MLCommons is joining Partnership on AI! 🚀 Better AI starts with better benchmarks. Together, we'll connect our open engineering community with a global network working toward the same goal: AI systems that are accurate, safe, and accountable. bit.ly/4t5usu9

We’re thrilled to announce we’ve joined @PartnershipAI, a multi-stakeholder community addressing the most important and difficult questions concerning the future of AI.

Want to explore the MLPerf Inference v6.0 results yourself? Dive into our interactive dashboard - filter by benchmark, system, and scenario to see how the latest hardware stacks up. 📊 🔗 bit.ly/3PLbCJR #MLPerf #MLCommons #AI #inference

MLPerf® Endpoints uses step functions, not trend lines. Interpolating between measured points can hide real failures: memory overflows, P99 spikes. Only verified operating points. No paper performance. mlcommons.org/2026/03/mlperf… #MLPerf #AIBenchmarking

What's new in MLPerf Inference v6.0? Watch the press briefing for a walkthrough of five new benchmarks, standout results, and what it all means for the future of AI inference. ▶️ 🔗 youtu.be/3FdkYZZlhDI #MLPerf #MLCommons #AI #inference

And a huge thank you to our partners @Meta, @Shopify, and @ultralytics for contributing real-world datasets and workloads that make these benchmarks truly production-representative.

@DaniellerSmci, @FollowStevens, and @Supermicro A special welcome to our first-time submitters: Inventec Corporation, @netwebtech , and @FollowStevens!

MLPerf Inference v6.0 is here - our most significant benchmark update ever. 5 new/updated benchmarks. 24 submitting organizations. Industry-first tests for text-to-video and speculative decoding. Full results: mlcommons.org/2026/04/mlperf… #MLPerf #MLCommons

MLPerf® Endpoints visualizes results as Pareto curves — not single-point benchmarks. See the full trade-off: throughput vs. interactivity, TTFT vs. latency. Designed for real procurement decisions. Explore early results: bit.ly/4bVlIiD #MLPerf #AIInfrastructure

One day out. MLPerf Inference v6.0 launches tomorrow, April 1. Why this one's different: new benchmarks that reflect where AI is actually headed - text-to-video, speculative decoding, vision-language models, and more. Results drop at 8 AM PDT. #MLPerf #MLCommons #AI

MLPerf Mobile is now available on Apple iOS and Android. From your phone or tablet, test ML performance across image classification, language understanding, super resolution, and text-to-image generation. Apple: apple.co/4uGhvIa Android: bit.ly/3NzTaTP

AI agents need datasets that describe themselves. Croissant 1.1 adds machine-actionable provenance, vocabulary interoperability, and embedded governance to 700K+ ML datasets. The agent-ready data standard is here: bit.ly/3NZk5bK #Croissant #MLCommons

MLPerf Inference v6.0 drops next Wednesday. New benchmarks for text-to-video, speculative decoding, and VLMs. Are you ready? #MLPerf

New in #MLPerf Inference v6.0: 🧠 GPT-OSS 120B — new open-weight MoE benchmark for math, coding & science reasoning ⚡ DeepSeek-R1 Interactive — low-latency scenario + the first #MLPerf standard for speculative decoding bit.ly/4m19oCl #MLCommons #LLMBenchmark #GenAI

AI safety certification shouldn't be a self-assessment. The AILuminate Global Assurance Program from MLCommons gives organizations rigorous, independent validation — built on the open AILuminate safety benchmark. bit.ly/4kIS18x

Measuring GenAI performance is harder than it looks. Accuracy + latency + throughput + sequence length = a non-linear, multi-dimensional surface. Simple scenarios miss it entirely. #MLPerf Endpoints is being designed to capture production reality. bit.ly/3Pjx34u

6/ Rolling submissions are planned for Q2 2026. Read more about what we're developing: 📖 mlcommons.org/2026/03/mlperf… Try it out here: Endpoints.MLCommons.org #MLPerf #GenerativeAI #AIBenchmarking #MLCommons #GTC2026

5/ We're sharing early demonstration results from @AMD , @Google , @intel , KRAI, and @nvidia — across models including DeepSeek-R1, Llama 3.1 8B, GPT OSS 120B, and QWEN 3 Coder 480B — with 30+ organizations already supporting the effort.

1/ At GTC, @MLCommons co-founder David Kanter (@TheKanter) walked through what we're developing with MLPerf® Endpoints — a new approach to benchmarking generative AI services built for how they're actually deployed. 🧵 mlcommons.org/2026/03/mlperf…