{"id":53827,"date":"2026-06-16T11:13:06","date_gmt":"2026-06-16T03:13:06","guid":{"rendered":"https:\/\/www.1ai.net\/?p=53827"},"modified":"2026-06-16T11:13:06","modified_gmt":"2026-06-16T03:13:06","slug":"%e5%b7%b4%e8%a5%bf%e6%a8%a1%e5%9e%8b-rio-3-5-%e5%bc%95%e5%8f%91%e3%80%8c%e7%bc%9d%e5%90%88%e3%80%8dqwen-%e8%b4%a8%e7%96%91","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/53827.html","title":{"rendered":"Brazilian model Rio 3.5 triggers \"suture\" Qwen challenge"},"content":{"rendered":"<p>June 16th<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e5%b7%b4%e8%a5%bf\" title=\"_Other Organiser\" target=\"_blank\" >Brazil<\/a>IT CORPORATION UNDER THE UMBRELLA OF THE MUNICIPALITY OF RIO DE JANEIRO<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e5%bc%80%e6%ba%90\" title=\"[View articles tagged with [open source]]\" target=\"_blank\" >Open Source<\/a>Rio 3.5 Open 397B <a href=\"https:\/\/www.1ai.net\/en\/tag\/%e6%a8%a1%e5%9e%8b\" title=\"_Other Organiser\" target=\"_blank\" >Model<\/a>In the AI community to trigger discussion. The model page shows that it is open in the MIT protocol and that the total parameter is about 397B, single token activates about 17B and supports a longer context\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-53828\" title=\"9d8720c0j00tpfkb003yd000ou00tsm\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2026\/06\/9d8720c0j00tgpfkb003yd000ou00tsm.jpg\" alt=\"9d8720c0j00tpfkb003yd000ou00tsm\" width=\"894\" height=\"1072\" \/><\/p>\n<p>However, the model was published less than 24 hours ago, and the open-source AI team, Nex-AGI, spoke on platform X, pointing out that Rio 3.5 was essentially a shell product made up of model consolidation (Model Merge) technology\u3002<\/p>\n<p>After an in-depth mathematical analysis of Rio 3.5 weights, Nex-AGI disclosed that the bottom formulation of the model was approximately Nex N2 Pro of 60% and Ali Baba Qwen 3.5 of 40%. The team provided two independent verification paths:<\/p>\n<p>One is that after removing the hard-coded \"you are Rio\" system hint, the deployed model has a 79% probability of claiming \"Nex from Nex-AGI\", a zero-probability for \"Rio\" and recapitulating Nex-AGI's customized background stories without a word<\/p>\n<p>The second is Rio 3.5 per weight of each of the 60-storey networks, which displays a fully consistent combination of 0.6.0.4 Nex and Qwen, with statistical deviations of up to thousands of standard deviations, which cannot be explained by regular fine-tuning\u3002<\/p>\n<p>The team stated that the presence of Rio 3.5 objectively reversed the capacity of the Nex open source base, but at the same time seriously emphasized that the signature and the acknowledgement were the limit that the open source community could not cross\u3002<\/p>","protected":false},"excerpt":{"rendered":"<p>On June 16th, a Rio 3.5 Open 397B model under the umbrella of the municipality of Rio de Janeiro, Brazil, has recently been the subject of discussion in the AI community. The model page shows that it is open in the MIT protocol, with a total parameter of about 397B, single token activates about 17B and supports a longer context. However, the model was published less than 24 hours ago, and the open-source AI team, Nex-AGI, spoke on platform X, pointing out that Rio 3.5 was essentially a shell product made up of model consolidation (Model Merge) technology. Nex-AGI, after in-depth mathematical analysis of Rio 3.5 weights, disclosed that the bottom formulation of the model was about 6<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[3608,219,1489],"collection":[],"class_list":["post-53827","post","type-post","status-publish","format-standard","hentry","category-news","tag-3608","tag-219","tag-1489"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/53827","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=53827"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/53827\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=53827"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=53827"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=53827"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=53827"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}