{"id":370686,"date":"2026-08-19T09:32:32","date_gmt":"2026-08-19T08:32:32","guid":{"rendered":"https:\/\/droix.net\/blogs\/?p=365004"},"modified":"2026-09-26T20:45:44","modified_gmt":"2026-09-26T19:45:44","slug":"geekom-four-node-ai-cluster","status":"publish","type":"post","link":"https:\/\/droix.net\/blogs\/en-au\/geekom-four-node-ai-cluster\/","title":{"rendered":"GEEKOM Demonstrates Four-Node Mini PC AI Cluster with Local DeepSeek Model"},"content":{"rendered":"\n

GEEKOM<\/a> has demonstrated a four-node AI cluster built from its compact A9 Mega Mini PCs, showing how small-form-factor systems can be combined for local enterprise AI applications.<\/p>\n\n\n\n

The cluster connects four A9 Mega<\/a> units via USB4, allowing them to operate as a distributed platform without requiring a traditional server rack or proprietary high-speed networking equipment. Each system is powered by AMD\u2019s Ryzen AI Max+ 395 <\/a>processor, featuring 16 Zen 5 CPU cores, Radeon 8060S graphics and unified memory.<\/p>\n\n\n\n

\n
\"Four<\/a>
Four A9 Mega units<\/figcaption><\/figure>\n\n\n\n
\"Four<\/a>
Four A9 Mega units<\/figcaption><\/figure>\n<\/figure>\n\n\n\n

Running Ubuntu, ROCm and DwarfStar software, the cluster distributes the DeepSeek V4 Flash model across all four machines. It also provides an OpenAI-compatible API, enabling integration with applications, AI agents and automation tools.<\/p>\n\n\n\n

A key advantage of the setup is local processing. Businesses can analyse documents, source code, contracts and internal knowledge without sending sensitive information to a public cloud. GEEKOM says the system has also been tested with context windows of up to 250,000 tokens, making it suitable for lengthy documents and complex workflows.<\/p>\n\n\n\n

\n
\"Four<\/a>
Four GEEKOM A9 Mega units<\/figcaption><\/figure>\n\n\n\n
\"Four<\/a>
Four A9 Mega units back view<\/figcaption><\/figure>\n<\/figure>\n\n\n\n

According to the company, the cluster achieved approximately 14.61 tokens per second at single concurrency, with a 95th-percentile time to first token of around 0.42 seconds in its 32- and 128-token tests. Organisations can begin with one or two A9 Mega<\/a> systems and expand to four as their requirements grow.<\/p>\n\n\n\n

The demonstration highlights the potential of compact PCs as scalable local AI platforms for businesses, laboratories, classrooms and edge-computing environments.<\/p>\n\n\n\n

Source: GEEKOM via PR Newswire<\/a>. Performance figures and capabilities are based on company-reported testing.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

GEEKOM has demonstrated a four-node AI cluster built from its compact A9 Mega Mini PCs. Connected via USB4 and powered by AMD Ryzen AI Max+ 395 processors, the system runs DeepSeek V4 Flash locally, offering businesses a scalable way to process sensitive documents, code and AI workloads without relying entirely on cloud infrastructure.<\/p>\n","protected":false},"author":859,"featured_media":370687,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6593,8491],"tags":[37353,8743,6597],"class_list":["post-370686","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","category-mini-pc-news","tag-a9-mega-27","tag-geekom","tag-mini-pc"],"lang":"en-au","translations":{"en-au":370686,"en":365004,"de":370301,"en-gb":370672,"en-ie":370679,"fr":371387,"ar":382327,"bg":384272,"cs":385446,"es":387053,"da":387374,"pt-pt":388137,"hr":389905,"hu":390629,"lv":393581,"el":403888,"et":403954,"fi":404020,"it":404085,"lt":404150,"nl":404259,"no":404325,"pl":404390,"ro":404456,"sk":404522,"sl":404588,"sv":404653},"pll_sync_post":{},"_links":{"self":[{"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/posts\/370686","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/users\/859"}],"replies":[{"embeddable":true,"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/comments?post=370686"}],"version-history":[{"count":1,"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/posts\/370686\/revisions"}],"predecessor-version":[{"id":370692,"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/posts\/370686\/revisions\/370692"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/media\/370687"}],"wp:attachment":[{"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/media?parent=370686"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/categories?post=370686"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/droix.net\/blogs\/wp-json\/wp\/v2\/tags?post=370686"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}