{"id":422488,"date":"2026-01-21T00:25:14","date_gmt":"2026-01-21T00:25:14","guid":{"rendered":"https:\/\/www.newsbeep.com\/ca\/422488\/"},"modified":"2026-01-21T00:25:14","modified_gmt":"2026-01-21T00:25:14","slug":"china-telecom-develops-countrys-first-moe-models-trained-entirely-on-huaweis-ai-chips","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ca\/422488\/","title":{"rendered":"China Telecom develops country\u2019s first MoE models trained entirely on Huawei\u2019s AI chips"},"content":{"rendered":"<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">This marked the first time a Chinese developer publicly validated the feasibility of using only Huawei chips to train AI models with MoE architecture, which has become widely adopted because of its ability to deliver high performance with fewer computational resources.<\/p>\n<p>The TeleChat3 models \u2013 ranging from 105 billion to trillions of parameters \u2013 were trained on Huawei\u2019s <a target=\"_self\" class=\"e1yy41x40 ef9u0v01 css-1ankfgb ecgc78b0\" href=\"https:\/\/www.scmp.com\/tech\/big-tech\/article\/3265640\/huawei-says-its-ai-chip-better-nvidias-a100-amid-chinas-self-reliance-drive?module=inline&amp;pgtype=article?module=inline&amp;pgtype=article\" title=\"\" data-qa=\"BaseLink-renderAnchor-StyledAnchor\" rel=\"nofollow noopener\">Ascend 910B<\/a> chips and its open-source deep learning AI framework MindSpore, according to a technical paper published last month by China Telecom\u2019s Institute of Artificial Intelligence (TeleAI).<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">The Huawei stack was able to meet the \u201csevere demands\u201d of training large-scale MoE models across many different sizes, TeleAI researchers said.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">\u201cThese contributions collectively address critical bottlenecks in frontier-scale model training, establishing a mature full-stack solution tailored to domestic computational ecosystems,\u201d they said.<\/p>\n<p>The initiative by China Telecom, one of the world\u2019s largest fixed-line and mobile network operators, underscored growing efforts by a number of US-blacklisted Chinese firms \u2013 including Huawei and <a target=\"_self\" class=\"e1yy41x40 ef9u0v01 css-1ankfgb ecgc78b0\" href=\"https:\/\/www.scmp.com\/tech\/big-tech\/article\/3314014\/tech-war-iflytek-says-using-local-chips-ai-models-extends-development-time-3-months?module=inline&amp;pgtype=article\" title=\"\" data-qa=\"BaseLink-renderAnchor-StyledAnchor\" rel=\"nofollow noopener\">iFlytek<\/a> \u2013 to train AI models using domestically designed <a target=\"_self\" class=\"e1yy41x40 ef9u0v01 css-1ankfgb ecgc78b0\" href=\"https:\/\/www.scmp.com\/topics\/semiconductors?module=inline&amp;pgtype=article\" title=\"\" data-qa=\"BaseLink-renderAnchor-StyledAnchor\" rel=\"nofollow noopener\">semiconductors<\/a>.<img decoding=\"async\" alt=\"Huawei Technology\u2019s Ascend 910B chip. Photo: Handout\" data-qa=\"BaseImage-handleRenderImage-StyledImage\" class=\"e1gf69pb2 css-6ikqhs e445x7d0\" loading=\"lazy\" src=\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2026\/01\/a08be7ea-27f1-485d-9c28-dacc3cb87d13_8f53a9eb.jpg\" title=\"Huawei Technology\u2019s Ascend 910B chip. Photo: Handout\"\/>Huawei Technology\u2019s Ascend 910B chip. Photo: Handout<\/p>\n","protected":false},"excerpt":{"rendered":"This marked the first time a Chinese developer publicly validated the feasibility of using only Huawei chips to&hellip;\n","protected":false},"author":2,"featured_media":422489,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[33153,62,19089,32732,110825,276,277,176953,8580,49,48,95,17115,176954,11119,143326,8051,176949,176955,176948,176952,6494,278,36103,61,176951,176950,12154,182,121207],"class_list":["post-422488","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-advanced-micro-devices","tag-ai","tag-ai-models","tag-alibaba-group-holding","tag-ant-group","tag-artificial-intelligence","tag-artificialintelligence","tag-ascend-910b-chips","tag-beijing","tag-ca","tag-canada","tag-china","tag-china-telecom","tag-deepseeks-v3-model","tag-fintech","tag-h200","tag-huawei","tag-iflytek","tag-mindspore","tag-mixture-of-experts-architecture","tag-moe-architecture","tag-nvidia","tag-openai","tag-semiconductors","tag-technology","tag-teleai","tag-telechat3-models","tag-tsinghua-university","tag-us","tag-zhipu-ai"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/422488","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/comments?post=422488"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/422488\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media\/422489"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media?parent=422488"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/categories?post=422488"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/tags?post=422488"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}