{"id":46450,"date":"2025-11-25T19:05:38","date_gmt":"2025-11-25T11:05:38","guid":{"rendered":"https:\/\/www.1ai.net\/?p=46450"},"modified":"2025-11-25T19:05:38","modified_gmt":"2025-11-25T11:05:38","slug":"%e6%96%b0%e5%8a%a0%e5%9d%a1%e5%9b%bd%e5%ae%b6-ai-%e8%ae%a1%e5%88%92%e6%94%be%e5%bc%83-meta-%e6%a8%a1%e5%9e%8b%ef%bc%8c%e8%bd%ac%e5%90%91%e9%98%bf%e9%87%8c%e5%8d%83%e9%97%ae","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/46450.html","title":{"rendered":"Singapore National AI plans to abandon the Meta model and turn to Aliwan"},"content":{"rendered":"

The news of November 25thAlibaba<\/a>Aliyun of the group announced on November 24th that it would beSingapore<\/a>The latest version of the \"Sea-Lion\" Large Language Model (LLM) of the National Artificial Intelligence Programme (AI Singapore, ASG) provides technical support. It's worth mentioning that ASG gave up this time Meta<\/a> Model, to Ali BabaThousand Questions on Tongyi<\/a> Qwen Open Source Architecture\u3002<\/p>\n

\"534<\/p>\n

In a press release, Aliyun stated that the release of the latest version of the Qwen3-32B basic model, based on Aliyun, marked a \u201csignificant step\u201d in promoting stronger and more inclusive artificial intelligence solutions in the region\u3002<\/p>\n

In addition, cooperation between the two sides included Aliyun providing technical support for the \u201cadvanced post-training phase\u201d of the larger model. Aliyun stressed that the basic model Qwen3-32B had received additional training of over 100 billion South-East Asian linguistics (tokens), which had significantly enhanced its ability to understand local expressions, dialogue nuances and regional areas of knowledge\u3002<\/p>\n

As the latest generation of the Qwen series, the Qwen3 Foundation model itself has completed pre-training on large and highly diverse data sets covering 119 languages and dialects, with a total of 36 trillion words. This has enabled it to have a broader range of language coverage from the initial stage and has strengthened support for the South-East Asian languages, which are usually underrepresented in mainstream artificial intelligence models\u3002<\/p>\n

As another key element of the cooperation, AI Singhapore contributed to its open-source regional data sets - including data collation, optimization and assessment for South-East Asian language missions\u3002<\/p>\n

At present, Qwen-Sea-Lion-v4 is ranked first on the South-East Asian Language Model Integrated Assessment (SALMA) list among large open source models with parameters of less than 200 billion. The model has been downloaded free of charge on the AI Singapore official network and the Hugging Face platform\u3002<\/p>\n

IT IS WORTH MENTIONING THAT IN DECEMBER 2023, SINGAPORE OFFICIALLY LAUNCHED A NATIONAL PROGRAMME TOTALLING NZ$70 MILLION (NOTE: THE CURRENT EXCHANGE RATE IS APPROXIMATELY NZ$381 MILLION) AIMED AT COMPREHENSIVELY UPGRADING THE COUNTRY ' S SCIENTIFIC AND ENGINEERING CAPABILITIES IN THE AREA OF MULTI-MODULAR MEGA-LINGUISTIC MODELS. THE PROJECT IS FUNDED BY THE SINGAPORE NATIONAL RESEARCH FOUNDATION (NRF) AND ENDORSED BY THE INFORMATION AND COMMUNICATION MEDIA DEVELOPMENT AUTHORITY (IMDA)\u3002<\/p>","protected":false},"excerpt":{"rendered":"

On November 25th, Ariyun, under the umbrella of the Ali Baba Group, announced on November 24 that technical support would be provided for the latest version of the \u201cSea-Lion\u201d Large Language Model (LLM), known as Qwen-Sea-Lion-v4, of the Singapore National Artificial Intelligence Programme (AI Singapore, ASG). It's worth mentioning that the AISG this time abandoned the Meta model to the Qwen Open Source Architecture of Ali Baba. In a press release, Aliyun stated that the release of the latest version of the Qwen3-32B basic model, based on Aliyun, marked the \"heavy\" of AISG in promoting stronger, more inclusive artificial intelligence solutions in the region<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[297,2496,331,390],"collection":[],"class_list":["post-46450","post","type-post","status-publish","format-standard","hentry","category-news","tag-meta","tag-2496","tag-331","tag-390"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/46450","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=46450"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/46450\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=46450"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=46450"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=46450"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=46450"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}