--- 格式版本: 2 标题: "NVIDIA announces Groq 3 LPX production for inference workloads" 原文链接: "https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/" 发布日期: "2026-09-01" 发布时间校准状态: "found" 发布时间需复核: "否" 发布时间来源: "llm:local:strict_original_body" 发布时间证据: "September 1, 2026" 发布时间校准原因: "标题附近明确标注的发布时间为September 1, 2026，符合文章发布时间。" 发布时间校准置信度: "1" 发布时间候选数量: 12 发布时间严格候选数量: 2 发布时间原页读取状态: "source template page reused from URL open" 发布时间未找到原因: "" 发布时间校准时间: "2026-09-02T01:44:43+08:00" 发布时间仲裁状态: "confirmed" 发布时间仲裁尝试次数: 1 发布时间仲裁耗时毫秒: 6435 发现时间: "2026-09-02T01:43:40+08:00" 入库时间: "2026-09-01T17:45:03.752Z" 来源平台: "Engineering.com 搜索" 搜索渠道: "source_template" 搜索词: "site:engineering.com Vera Rubin" 匹配关键词: - "NVL72" - "Vera Rubin" - "deployment" - "performance" - "latency" - "throughput" - "AI" 相关厂家: - "NVIDIA" 相关专家: [] 内容类型: "网页" 抓取工具: "Jina Reader" 清洗工具: "Jina Reader Markdown + Defuddle/Readability 正文提取" 原始附件: [] AI摘要: "NVIDIA宣布其AI推理加速器Groq 3 LPX进入全面量产，该产品属于Vera Rubin平台，旨在提升智能体AI场景的token生成速度。" AI摘要模型: "ali-deepseek-v4-flash" AI摘要时间: "2026-09-01T23:06:48.148Z" 人工打分: "是" 人工更新时间: "2026-09-02T15:46:33+08:00" AI优质: "否" AI打分: 54 AI分档: "非优质" AI质检状态: "不通过" AI打分理由: "正文主线是Groq 3 LPX推理加速器量产及其与Vera Rubin NVL72机架平台的配合，给出3400 tokens/s、Gemma 4 31B、10万token上下文、七芯片五类机架及BlueField-4、Spectrum-6等信息，并披露Nebius部署计划。来源为工程媒体对NVIDIA公告的二手改写，缺少原始公告链接和独家分析。历史库已存在同一事件的NVIDIA官方原文，量产、性能及Nebius首批采用等核心事实均已完整出现，本文没有可核验新增，当前页面自身不具备优先保留价值，命中重复二手稿强否决，封顶54分。" AI质检模型: "gpt-5.6-sol" AI质检时间: "2026-09-02T09:45:09+08:00" AI主题相关性: 17 AI来源权威性: 8 AI新颖性: 2 AI技术细节: 11 AI商业部署信号: 8 AI完整性: 8 AI评分提示词版本: "v17-精简生产版" AI评分提示词SHA256: "48fb9777f386026761b4873eaff30807694fb11e9b352d7c69bf2dfde750cc7d" AI评分知识库版本: "knowledge_base_v1-20260819+runtime.81" AI评分知识库SHA256: "3b93d12e47749b3512f545f51c44c011bdc0931677c2cfe61e4df59a2b1a5a48" AI评分知识库检索词: "[\"NVIDIA\",\"Vera Rubin\",\"site:engineering.com Vera Rubin\",\"NVL72\",\"Rubin\",\"rack-scale\",\"RAS\",\"DPU\",\"Intel\",\"LPX\",\"URL\",\"IT\"]" AI评分知识库命中: "[{\"id\":\"july-correct-0020\",\"title\":\"NVIDIA Vera Rubin：引領代理 AI 的時代\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"NVIDIA\",\"Vera Rubin\",\"NVL72\",\"Rubin\",\"RAS\",\"DPU\",\"LPX\"],\"rank\":-24.43018848675334},{\"id\":\"runtime-d00d6246caee468c0053ce6f\",\"title\":\"NVIDIA Groq 3 LPX Now in Full Production With World-Class Speed for Agentic AI\",\"sourceType\":\"ai_excellent_article\",\"time\":\"\",\"matchedTerms\":[\"NVIDIA\",\"Vera Rubin\",\"NVL72\",\"Rubin\",\"rack-scale\",\"RAS\",\"DPU\",\"Intel\",\"LPX\",\"IT\"],\"rank\":-22.106365674327314},{\"id\":\"july-correct-0034\",\"title\":\"AMD Fires Back at Nvidia with Helios AI System, Epyc CPUs\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"NVIDIA\",\"Vera Rubin\",\"NVL72\",\"Rubin\",\"rack-scale\",\"RAS\",\"DPU\",\"Intel\",\"URL\",\"IT\"],\"rank\":-18.05747487510997},{\"id\":\"july-correct-0095\",\"title\":\"Inside NVIDIA Rubin GPU Architecture: Powering the Era of Agentic AI\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"NVIDIA\",\"Vera Rubin\",\"NVL72\",\"Rubin\",\"rack-scale\",\"RAS\",\"Intel\",\"URL\",\"IT\"],\"rank\":-15.687405303861023},{\"id\":\"july-correct-0017\",\"title\":\"AMD challenges Nvidia’s networking dominance with Helios racks boasting 50% higher bandwidth\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"NVIDIA\",\"Vera Rubin\",\"NVL72\",\"Rubin\",\"rack-scale\",\"RAS\",\"DPU\",\"URL\",\"IT\"],\"rank\":-15.52540581964668}]" 采集批次: "2026年9月1日23点56分52秒" 采集批次ID: "20260901-235652-761" 去重键: "https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads" --- Title: NVIDIA announces Groq 3 LPX production for inference workloads URL Source: https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/ Published Time: 2026-08-25T11:04:01+00:00 Markdown Content: [Skip to content](https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/#main-content) Search [Subscribe](https://www.engineering.com/subscribe/) [Technology](https://www.engineering.com/category/technology/) Back Technology [Browse All All](https://www.engineering.com/category/technology/) [3D Printing ](https://www.engineering.com/category/technology/3d-printing/)[Advanced Manufacturing ](https://www.engineering.com/category/technology/advanced-manufacturing/)[Artificial Intelligence ](https://www.engineering.com/category/technology/artificial-intelligence/)[Automation ](https://www.engineering.com/category/technology/automation/)[BIM ](https://www.engineering.com/category/technology/bim/)[Components ](https://www.engineering.com/category/technology/components/)[Computing ](https://www.engineering.com/category/technology/computing/)[Design ](https://www.engineering.com/category/technology/design/)[Digital Transformation ](https://www.engineering.com/category/technology/digital-transformation/)[Off Topic ](https://www.engineering.com/category/technology/off-topic/)[PLM/ERP ](https://www.engineering.com/category/technology/plm-erp/)[Simulation ](https://www.engineering.com/category/technology/simulation/) [Industry](https://www.engineering.com/category/industry/) Back Industry [Browse All All](https://www.engineering.com/category/industry/) [Aerospace and Defense ](https://www.engineering.com/category/industry/aerospace-and-defense/)[Automotive ](https://www.engineering.com/category/industry/automotive/)[Education ](https://www.engineering.com/category/industry/education/)[Energy ](https://www.engineering.com/category/industry/energy/)[Industry News ](https://www.engineering.com/category/industry/industry-news/)[Quality ](https://www.engineering.com/category/technology/quality/) ![Image 20](./assets/img-9c69699e.jpg) [3D Printing](https://www.engineering.com/category/technology/3d-printing/) ### [3D-Printed Swag Ideas for Your Next Trade Show](https://www.engineering.com/3d-printed-swag-ideas-for-your-next-trade-show/) ![Image 21](./assets/img-9196cce4.jpg) [Advanced Manufacturing](https://www.engineering.com/category/technology/advanced-manufacturing/) ### [Siemens and Nvidia building next-gen ‘industrial AI operating system’](https://www.engineering.com/siemens-and-nvidia-building-next-gen-industrial-ai-operating-system/) [Resources](https://www.engineering.com/resources/) Back Resources [Resources All](https://www.engineering.com/resources/) [Papers, Reports, and more Research on the latest technology and processes prepared by domain experts.](https://www.engineering.com/resources/)[Calculators Specialty calculators for engineering professionals.](https://www.engineering.com/resources/?_resource_category=calculators)[Webinars Discussion on technology and trends hosted by engineering leaders.](https://gateway.on24.com/wcc/experience/elitewtwhmedia/2927718/4498137/eng)[Voices Explore insights from company thought leaders as they share their expertise and industry knowledge.](https://www.engineering.com/voices/)[Views Discover company news, insights, case studies, testimonials, and announcements from a variety of organizations.](https://www.engineering.com/views/) [Communities](https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/#) Back Communities [Eng-Tips Forums for engineering.com professionals](https://www.eng-tips.com/)[Tek-Tips Forums for IT professionals](https://www.tek-tips.com/)[Engineersrule Solidworks news and tips](https://www.engineersrule.com/)[EDA Board Forums for electronic design professionals](https://www.edaboard.com/)[Electro Tech Forums for building electronic circuits, projects and gadgets.](https://www.electro-tech-online.com/) [Stem Games](https://www.engineering.com/games/) Back Stem Games [Browse all All](https://www.engineering.com/games/) Problem solving games to challenge engineers (or help them procrastinate). [Block Blast Who needs Tetris when you have Block Blast?](https://www.engineering.com/games/block-blast/)[Temple Run 2 Take the idol... if you dare.](https://www.engineering.com/games/temple-run-2/)[2048 It's exponential fun!](https://www.engineering.com/games/2048/) [TV](https://www.engineering.com/category/watch) Back TV [Browse All All](https://www.engineering.com/category/watch) [This Week in Engineering ](https://www.engineering.com/category/watch/this-week-in-engineering/)[End of the Line ](https://www.engineering.com/category/watch/end-of-the-line/)[The Engineering Roundtable ](https://www.engineering.com/category/watch/the-engineering-roundtable/)[Designing the Future ](https://www.engineering.com/category/watch/designing-the-future/)[Manufacturing the Future ](https://www.engineering.com/category/watch/manufacturing-the-future/)[The Primary Loop ](https://www.engineering.com/category/watch/the-primary-loop/)[On the Floor ](https://www.engineering.com/category/watch/on-the-floor/) [Careers](https://www.engineering.com/category/industry/careers/) Back Careers [Browse All All](https://www.engineering.com/category/industry/careers/) [EDI ](https://www.engineering.com/category/edi/)[Education ](https://www.engineering.com/category/industry/education/)[Top Workplaces ](https://www.engineering.com/category/industry/top-workplaces/) [Top Workplaces](https://www.engineering.com/category/industry/top-workplaces/) ### [These are the Top Workplaces for Engineers in 2026](https://www.engineering.com/these-are-the-top-workplaces-for-engineers-in-2026/) Aug 25, 2026 [Industry News](https://www.engineering.com/category/industry/industry-news/) # NVIDIA announces Groq 3 LPX production for inference workloads Benchmark data showed 3,400 tokens/sec on Gemma 4 31B with a 100,000-token context window. ![Image 46](./assets/img-093db0d6.webp) NVIDIA announced that NVIDIA Groq 3 LPX, an AI inference accelerator, is now in full production. As part of the NVIDIA Vera Rubin platform, Groq 3 LPX is designed to increase token generation speed for agentic AI systems. Agentic systems often generate large token volumes across many inference steps, so token generation speed affects how quickly agents can process tasks. NVIDIA positions Groq 3 LPX to support faster response times while handling context-heavy workloads. NVIDIA also states that Vera Rubin NVL72 systems support both training and inference in AI factory environments. Groq 3 LPX is intended to extend inference performance on Vera Rubin NVL72 by increasing per-user token generation rates. NVIDIA reported benchmark results from Artificial Analysis showing 3,400 output tokens per second on Gemma 4 31B with a 100,000-token context window. The company describes this as a reference point for high-context agentic inference performance. NVIDIA says Groq 3 LPX can reduce completion time for some agentic tasks, including coding workflows, and improve responsiveness in latency-sensitive use cases compared with alternative platforms. ### Groq 3 LPX for interactive inference Agentic AI creates two main compute requirements: processing large context windows efficiently and generating tokens with low latency. NVIDIA Groq 3 LPX is designed to improve interactivity by increasing token generation speed for individual users. Faster token generation can allow agents to complete more steps in a given time, such as checking files, writing and testing code, calling tools and validating outputs while maintaining responsive interaction. ### AI cloud deployment plans for Groq 3 LPX AI cloud providers are expanding infrastructure for training, reasoning and inference at scale. NVIDIA positions Groq 3 LPX as an option for providers that run high-volume, latency-sensitive inference workloads in rack-scale deployments. Nebius plans to deploy NVIDIA Groq 3 LPX in Nebius Token Factory, its production inference platform, to provide developers with higher token generation throughput for agentic AI applications. Groq also plans to be an early adopter of the platform after Nebius. ### Co-designed AI factory platform NVIDIA describes Vera Rubin as a co-designed platform spanning seven chips and five rack types. NVIDIA says Vera Rubin NVL72 and Groq 3 LPX are intended to address different workload requirements across AI factory deployments, including frontier model developers and open model service providers. These rack platforms include NVIDIA BlueField-4 DPUs and operate with NVIDIA Vera CPU racks, NVIDIA Vera BlueField-4 STX storage and NVIDIA Spectrum-6 SPX Ethernet. NVIDIA states this configuration is designed to improve throughput per watt and reduce inference latency for multi-agent systems. For more information, visit [nvidia.com](https://www.nvidia.com/en-us/). ## You might also like ![Image 47: NVIDIA introduces Rubin platform for large-scale AI systems](./assets/img-f4ed3e1f.webp) #### [NVIDIA introduces Rubin platform for large-scale AI systems](https://www.engineering.com/nvidia-introduces-rubin-platform-for-large-scale-ai-systems/) [Read](https://www.engineering.com/nvidia-introduces-rubin-platform-for-large-scale-ai-systems/)![Image 48: Supermicro unveils AI systems based on NVIDIA Vera Rubin](./assets/img-86ed0d64.webp) #### [Supermicro unveils AI systems based on NVIDIA Vera Rubin](https://www.engineering.com/supermicro-unveils-ai-systems-based-on-nvidia-vera-rubin/) [Read](https://www.engineering.com/supermicro-unveils-ai-systems-based-on-nvidia-vera-rubin/)![Image 49: NVIDIA unveils Rubin CPX for massive-context inference](./assets/img-86590c53.webp) #### [NVIDIA unveils Rubin CPX for massive-context inference](https://www.engineering.com/nvidia-unveils-rubin-cpx-for-massive-context-inference/) [Read](https://www.engineering.com/nvidia-unveils-rubin-cpx-for-massive-context-inference/)![Image 50: NVIDIA introduces BlueField-4 platform for AI context memory](./assets/img-828a13bc.webp) #### [NVIDIA introduces BlueField-4 platform for AI context memory](https://www.engineering.com/nvidia-introduces-bluefield-4-platform-for-ai-context-memory/) [Read](https://www.engineering.com/nvidia-introduces-bluefield-4-platform-for-ai-context-memory/)![Image 51: NVIDIA ramps Vera Rubin for agentic AI infrastructure](./assets/img-82f33fd6.webp) #### [NVIDIA ramps Vera Rubin for agentic AI infrastructure](https://www.engineering.com/nvidia-ramps-vera-rubin-for-agentic-ai-infrastructure/) [Read](https://www.engineering.com/nvidia-ramps-vera-rubin-for-agentic-ai-infrastructure/) [ Written by #### Puja Mitra Puja Mitra has an MBA in Marketing and HR as well as an MA in Economics. As a Managing Editor, she has experience managing CAD, CAM, and CAE directories. She also handles design, BIM, manufacturing, digital transformation, and computing news. With over 14 years of editing experience, she has a particular interest in content and technical writing.](https://www.engineering.com/author/puja-mitra-2/) [Facebook](https://www.facebook.com/sharer/sharer.php?u=https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/&display=popup&ref=plugin&src=share_button "Share on Facebook")[Tweet](https://twitter.com/share?url=https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/&text=NVIDIA%20announces%20Groq%203%20LPX%20production%20for%20inference%20workloads "Share on Twitter")[LinkedIn](https://www.linkedin.com/shareArticle?mini=true&url=https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/ "Share on LinkedIn")[Reddit](https://www.reddit.com/submit?url=https://www.engineering.com/nvidia-announces-groq-3-lpx-production-for-inference-workloads/ "Share on Reddit") [ Written by #### Puja Mitra](https://www.engineering.com/author/puja-mitra-2/) ## more news [Industry News](https://www.engineering.com/category/industry/industry-news/) ### [Rockwell Automation adds 24/7 remote support assistance](https://www.engineering.com/rockwell-automation-adds-24-7-remote-support-assistance/) [Industry News](https://www.engineering.com/category/industry/industry-news/) ### [U.S. Navy funds Phase II for embedded-power UAV composites](https://www.engineering.com/u-s-navy-funds-phase-ii-for-embedded-power-uav-composites/) [Industry News](https://www.engineering.com/category/industry/industry-news/) ### [CollPlant to acquire LightSolver in photonic computing deal](https://www.engineering.com/collplant-to-acquire-lightsolver-in-photonic-computing-deal/) stay in the know ##### subscribe to the engineering.com ##### newsletter Δ Name This field is for validation purposes and should be left unchanged. This field is hidden when viewing the form Submission URL Email(Required) This field is hidden when viewing the form Primary Cat Subscribe Now ## related topics [Browse All](https://www.engineering.com/category/industry/) [Aerospace and Defense ](https://www.engineering.com/category/industry/aerospace-and-defense/)[Automotive ](https://www.engineering.com/category/industry/automotive/)[Careers ](https://www.engineering.com/category/industry/careers/)[Consumer Technology ](https://www.engineering.com/category/industry/consumer-technology/)[Education ](https://www.engineering.com/category/industry/education/)[Energy ](https://www.engineering.com/category/industry/energy/)[Industry News ](https://www.engineering.com/category/industry/industry-news/)[Top Workplaces ](https://www.engineering.com/category/industry/top-workplaces/) ## you might also like [Industry News](https://www.engineering.com/category/industry/industry-news/) ### [SWI Group joins NVIDIA Cloud Partner program](https://www.engineering.com/swi-group-joins-nvidia-cloud-partner-program-2/) September 1, 2026 [Read SWI Group joins NVIDIA Cloud Partner program](https://www.engineering.com/swi-group-joins-nvidia-cloud-partner-program-2/) ![Image 63](./assets/img-a11e1b8f.jpg) [Industry News](https://www.engineering.com/category/industry/industry-news/) ### [Innoviz appoints Yoav Har-Even as board chairperson](https://www.engineering.com/innoviz-appoints-yoav-har-even-as-board-chairperson/) September 1, 2026 [Read Innoviz appoints Yoav Har-Even as board chairperson](https://www.engineering.com/innoviz-appoints-yoav-har-even-as-board-chairperson/) ![Image 64](./assets/img-a4949329.jpg) [Industry News](https://www.engineering.com/category/industry/industry-news/) ### [SiMa.ai and ARK Electronics partner on autonomous drones](https://www.engineering.com/sima-ai-and-ark-electronics-partner-on-autonomous-drones/) September 1, 2026 [Read SiMa.ai and ARK Electronics partner on autonomous drones](https://www.engineering.com/sima-ai-and-ark-electronics-partner-on-autonomous-drones/) ![Image 65](./assets/img-32af095f.webp) [Industry News](https://www.engineering.com/category/industry/industry-news/) ### [Anyon and KMT partner on quantum computing deployment](https://www.engineering.com/anyon-and-kmt-partner-on-quantum-computing-deployment/) September 1, 2026 [Read Anyon and KMT partner on quantum computing deployment](https://www.engineering.com/anyon-and-kmt-partner-on-quantum-computing-deployment/) stay in the know Δ Email This field is for validation purposes and should be left unchanged. This field is hidden when viewing the form Category This field is hidden when viewing the form Submission URL Email(Required) Submit * [About](https://www.engineering.com/about/) * [Advertise](https://www.mediakit.engineering.com/) * [Contact](https://www.engineering.com/contact/) [](https://www.facebook.com/engineeringcom)[](https://www.instagram.com/engineeringdotcom)[](https://x.com/engineeringcom)[](https://www.linkedin.com/company/1844762/admin/feed/posts/)[](https://www.youtube.com/@engineeringdotcom) Part of the designworld digital network [Eng-Tips eng-tips.com](https://www.eng-tips.com/)[Design World designworldonline.com](https://www.designworldonline.com/)[Bearing Tips bearingtips.com](https://www.bearingtips.com/)[Coupling Tips couplingtips.com](https://www.couplingtips.com/)[Linear Motion Tips linearmotiontips.com](https://www.linearmotiontips.com/)[Motion Control tips motioncontroltips.com](https://www.motioncontroltips.com/)[The Robot Report therobotreport.com](https://www.therobotreport.com/)[Engineering Exchange engineeringexchange.com](https://engineeringexchange.com/)[3d CAD forums 3dcadforums.com](https://www.3dcadforums.com/)[Wire and Cable Tips wireandcabletips.com](https://www.wireandcabletips.com/)[Plant Engineering plantengineering.com](https://www.plantengineering.com/)[Control Engineering controleng.com](https://www.controleng.com/) © 2026 Arrowfly LLC. All Rights Reserved. * [Privacy Policy](https://www.arrowfly.com/privacy-policy/) * [Terms & Conditions](https://www.arrowfly.com/terms/) Notifications