{"id":11456,"date":"2026-08-26T14:45:26","date_gmt":"2026-08-26T06:45:26","guid":{"rendered":"https:\/\/infernews.com\/blog\/ai-cube-2000\/"},"modified":"2026-08-26T16:33:22","modified_gmt":"2026-08-26T08:33:22","slug":"ai-cube-2000","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/ai-cube-2000\/","title":{"rendered":"\u5c0f\u7c73 AI Cube \u4ee5\u4e09\u6b3e\u7384\u6212\u6676\u7247\u672c\u6a5f\u904b\u884c 2000 \u5104\u6a21\u578b"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/pasted-87a4f86198fd.jpg\" alt=\"Og image\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u5c0f\u7c73\u81ea 2021 \u5e74\u91cd\u555f\u5927\u6676\u7247\u7814\u767c\u8a08\u5283\u5f8c\uff0c\u5df2\u6295\u5165\u8d85\u904e 210 \u5104\u5143\u4eba\u6c11\u5e63\uff0c\u6676\u7247\u5718\u968a\u589e\u81f3\u8fd1 3,000 \u540d\u5c08\u5bb6\u3002\u5718\u968a\u5728\u7384\u6212\u6676\u7247\u6280\u8853\u6e9d\u901a\u6703\u5c55\u793a AI Cube \u5de5\u7a0b\u539f\u578b\uff0c\u5617\u8a66\u8655\u7406\u672c\u6a5f\u904b\u884c\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u6642\u7684\u904b\u7b97\u8207\u529f\u8017\u53d6\u6368\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">AI Cube \u7531\u7384\u6212 O3\u3001O100 \u53ca D100 \u4e09\u6b3e\u81ea\u7814\u6676\u7247\u5354\u540c\u5de5\u4f5c\u3002O3 \u8ca0\u8cac\u4e00\u822c\u8655\u7406\u8207\u5716\u50cf\u904b\u7b97\uff0cO100 \u91dd\u5c0d\u88dd\u7f6e\u7aef\u5927\u578b\u6a21\u578b\u63d0\u4f9b\u9ad8\u983b\u5bec\u52a0\u901f\uff0cD100 \u5247\u63d0\u4f9b\u9ad8\u7b97\u529b AI \u904b\u7b97\u53ca\u7d71\u4e00\u8a18\u61b6\u9ad4\uff0c\u6700\u9ad8\u652f\u63f4 2,000 \u5104\u53c3\u6578\u6a21\u578b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u539f\u578b\u6a5f\u6301\u7e8c\u529f\u8017\u6700\u9ad8 150W\uff0c\u4e26\u53ef\u90e8\u7f72 120B \u53ca 3B \u5169\u7a2e\u898f\u6a21\u6a21\u578b\uff0c\u5728\u5feb\u901f\u8207\u6162\u901f\u7cfb\u7d71\u4e4b\u9593\u5207\u63db\uff0c\u7528\u4f86\u9a57\u8b49\u4e0d\u540c\u6676\u7247\u7684\u5206\u5de5\u3002O100 \u7684\u8fd1\u8a18\u61b6\u9ad4\u904b\u7b97\u983b\u5bec\u6700\u9ad8\u9054 1.22TB\/s\uff0c\u5b98\u65b9\u8868\u793a\u88dd\u7f6e\u7aef\u5927\u578b\u6a21\u578b\u63a8\u7406\u901f\u5ea6\u6700\u9ad8\u53ef\u9054 330 TPS\u3002<\/p>\n\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"Xiaomi Just Shocked NVIDIA &mdash; A Tiny AI Cube Running 120B Locally\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL__YpREA0ph-k\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2F_YpREA0ph-k%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/_YpREA0ph-k\" \/><meta itemprop=\"duration\" content=\"PT10M5S\" \/><meta itemprop=\"uploadDate\" content=\"2026-08-25T11:45:43Z\" \/><\/div><meta itemprop=\"accessibilityFeature\" content=\"captions\" \/><div id=\"lyte__YpREA0ph-k\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2F_YpREA0ph-k%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">Xiaomi Just Shocked NVIDIA \u2014 A Tiny AI Cube Running 120B Locally<\/div><\/div><button tabindex=\"0\" class=\"play\"><\/button><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/_YpREA0ph-k\" rel=\"nofollow\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2F_YpREA0ph-k%2F0.jpg\" alt=\"Xiaomi Just Shocked NVIDIA &mdash; A Tiny AI Cube Running 120B Locally\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"Xiaomi AI Cube is a compact local AI computer designed to run a 120B model and a fast 3B model at the same time\u2014using Xiaomi\u2019s new O100 and D100 AI chips at up to 150 watts. Could this tiny local AI machine challenge NVIDIA DGX Spark and high-memory Mac Studio systems? In this video, we examine the Xiaomi AI Cube, its O100 AI accelerator, the D100 chip and Xiaomi\u2019s claim that a single chip can deploy models with more than 200 billion parameters. The most impressive number is memory bandwidth. Xiaomi says the O100 delivers up to 1.22 TB\/s through 3D wafer-level memory stacking\u2014higher on paper than the 819 GB\/s available on Apple\u2019s M3 Ultra. However, bandwidth alone does not prove that the AI Cube is faster. Xiaomi demonstrated the prototype running a 120B model alongside a smaller 3B MiMo model. The smaller model reportedly reached up to 330 tokens per second, but Xiaomi has not revealed the larger model\u2019s speed, quantization or context length. Chapters: 00:00 Xiaomi AI Cube Runs a 120B Model Locally 00:39 The Big Catch 01:07 Xiaomi O100 and D100 Explained 01:54 Xiaomi\u2019s Fast-and-Slow AI System 02:15 Why Local AI Hits the Memory Wall 02:43 Xiaomi\u2019s 3D Stacked Memory 03:12 O100 vs Apple M3 Ultra Bandwidth 03:34 330 Tokens per Second Explained 04:16 D100 and the 200B Model Claim 04:54 Can 200B Parameters Fit in 160GB? 05:28 What Xiaomi Has Not Revealed 05:52 Xiaomi AI Cube in Real-World Use 06:36 The Missing Software Ecosystem 07:29 Xiaomi AI Cube vs DGX Spark and Mac Studio 08:39 Why the Price Could Change Everything 09:05 Final Verdict: Did Xiaomi Kill DGX Spark? In this video: \u2022 Xiaomi AI Cube and its 120B local AI model \u2022 Xiaomi O100 and D100 AI chips explained \u2022 How the fast-and-slow dual-model system works \u2022 1.22 TB\/s near-memory bandwidth \u2022 160GB unified memory and the 200B model claim \u2022 Up to 330 tokens per second with a 3B model \u2022 Xiaomi AI Cube vs NVIDIA DGX Spark \u2022 Xiaomi AI Cube vs Apple M3 Ultra Mac Studio \u2022 Local coding, reasoning and AI-agent workloads \u2022 Missing price, software and release-date details \u2022 Whether Xiaomi has built a real DGX Spark alternative Important context: the Xiaomi AI Cube is currently an engineering prototype. Xiaomi has not announced its price, release date, operating system, supported model formats or the actual performance of its 120B model. The company also has not published the D100\u2019s memory bandwidth, power consumption or full AI performance. Supporting a 200B model does not necessarily mean running it quickly; that claim depends heavily on quantization, context length and software overhead. My verdict: the hardware concept is genuinely interesting. A compact 150-watt machine with up to 160GB of unified memory and 1.22 TB\/s of near-memory bandwidth could become a serious local AI platform. But until Xiaomi reveals the price, software stack and real 120B token speed, it is too early to call it a DGX Spark killer. Would you choose the Xiaomi AI Cube over an NVIDIA DGX Spark or a high-memory Mac Studio? Watch next: \u2022 This AI Model Is 100% Free Right Now \u2014 Unlimited Tokens, No Catch? https:\/\/youtu.be\/LCoA_guLBs0?si=l5tT4RC3zr9cWrUo \u2022 This New Engine Made an 8GB Laptop Run a 35B Model at 40 TPS https:\/\/www.youtube.com\/watch?v=MggRcCnFYRo&amp;t=14s Sources: \u2022 Reuters \u2014 Xiaomi\u2019s new Xring chips: https:\/\/www.reuters.com\/world\/china\/xiaomi-launches-new-xring-chip-partners-with-tsmc-production-sources-say-2026-08-24\/ \u2022 ZhiDongXi \u2014 O100, D100 and AI Cube technical details: https:\/\/zhidx.com\/p\/587490.html \u2022 IT Home \u2014 Xiaomi AI Cube prototype: https:\/\/m.ithome.com\/html\/993546.htm \u2022 Sina \/ MyDrivers \u2014 120B local model and 150W system: https:\/\/finance.sina.com.cn\/tech\/roll\/2026-08-24\/doc-inipkzxw2913585.shtml \u2022 CnEVPost \u2014 Xiaomi Xring D100: https:\/\/cnevpost.com\/2026\/08\/24\/xiaomi-unveils-xring-d100-chip\/ \u2022 NVIDIA DGX Spark specifications: https:\/\/www.nvidia.com\/en-us\/products\/workstations\/dgx-spark\/ \u2022 Apple Mac Studio specifications: https:\/\/www.apple.com\/mac-studio\/specs\/ Voice and synthesis: Narration voice derived from the CSTR VCTK Corpus, University of Edinburgh \u2014 CC BY 4.0. https:\/\/datashare.ed.ac.uk\/handle\/10283\/3443 Synthesized with Chatterbox by Resemble AI \u2014 MIT License. https:\/\/github.com\/resemble-ai\/chatterbox #Xiaomi #LocalAI #AICube\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n\n<p class=\"wp-block-paragraph\">\u5c0d\u9700\u8981\u672c\u6a5f\u8655\u7406 AI \u5de5\u4f5c\u8ca0\u8f09\u3001\u53c8\u5e0c\u671b\u6e1b\u5c11\u96f2\u7aef\u4f9d\u8cf4\u7684\u5de5\u4f5c\u6d41\uff0c\u9019\u7a2e\u591a\u6676\u7247\u8a2d\u8a08\u63d0\u4f9b\u53e6\u4e00\u7a2e\u65b9\u5411\uff0c\u4f46\u76ee\u524d\u4ecd\u8981\u7559\u610f\u539f\u578b\u6a5f\u8207\u6b63\u5f0f\u7522\u54c1\u4e4b\u9593\u7684\u5dee\u8ddd\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4ee5\u4e09\u6b3e\u7384\u6212\u6676\u7247\u5206\u5de5\u8655\u7406 CPU\u3001GPU\u3001NPU \u53ca\u5927\u578b\u6a21\u578b\u904b\u7b97<\/li>\n\n\n\n<li>\u6700\u9ad8\u652f\u63f4 2,000 \u5104\u53c3\u6578\u6a21\u578b\uff0c\u539f\u578b\u6a5f\u5df2\u5c55\u793a 120B \u8207 3B \u6a21\u578b\u914d\u7f6e<\/li>\n\n\n\n<li>\u6301\u7e8c\u529f\u8017\u6700\u9ad8 150W\uff0c\u4e26\u652f\u63f4\u5feb\u901f\u53ca\u6162\u901f\u7cfb\u7d71\u5207\u63db<\/li>\n\n\n\n<li>\u7384\u6212 O3 \u5df2\u5c55\u958b\u898f\u6a21\u91cf\u7522\uff0cO100 \u53ca D100 \u9810\u8a08 2027 \u5e74\u6b63\u5f0f\u5546\u7528<\/li>\n\n\n\n<li>\u5c0f\u7c73\u5c1a\u672a\u516c\u5e03 AI Cube \u7684\u4e0a\u5e02\u6642\u9593\u53ca\u552e\u50f9<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u5c0f\u7c73\u4ee5\u4e09\u6b3e\u81ea\u7814\u6676\u7247\u7d44\u6210 AI Cube \u5de5\u7a0b\u539f\u578b\uff0c\u5c07\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u904b\u7b97\u5e36\u5230\u672c\u6a5f\uff0c\u4f46\u8ddd\u96e2\u6b63\u5f0f\u4e0a\u5e02\u4ecd\u6709\u4e00\u6bb5\u8ddd\u96e2\u3002<\/p>\n","protected":false},"author":8,"featured_media":11455,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[206],"tags":[],"class_list":["post-11456","post","type-post","status-publish","format-standard","hentry","category-xiaomi"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11456","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=11456"}],"version-history":[{"count":2,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11456\/revisions"}],"predecessor-version":[{"id":11459,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11456\/revisions\/11459"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/11455"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=11456"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=11456"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=11456"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}