{"id":44203,"date":"2025-09-29T19:27:09","date_gmt":"2025-09-29T11:27:09","guid":{"rendered":"https:\/\/www.1ai.net\/?p=44203"},"modified":"2025-09-29T19:27:32","modified_gmt":"2025-09-29T11:27:32","slug":"deepseek-v3-2-exp-%e6%a8%a1%e5%9e%8b%e6%ad%a3%e5%bc%8f%e5%8f%91%e5%b8%83%e5%b9%b6%e5%bc%80%e6%ba%90%ef%bc%8capi-%e5%a4%a7%e5%b9%85%e9%99%8d%e4%bb%b7","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/44203.html","title":{"rendered":"The DeepSeek-V3.2-Exp model is officially published and is open, and API has significantly reduced prices"},"content":{"rendered":"<p>September 29th.<a href=\"https:\/\/www.1ai.net\/en\/tag\/deepseek\" title=\"[View articles tagged with [DeepSeek]]\" target=\"_blank\" >DeepSeek<\/a> Officially released today\u00a0<strong>DeepSeek-V3.2-Exp model<\/strong>It's an experimental version\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-44204\" title=\"e8f11504j00t3cl4i009ed000v90nlp\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2025\/09\/e8f11504j00t3cl4i009ed000v900nlp.jpg\" alt=\"e8f11504j00t3cl4i009ed000v90nlp\" width=\"1125\" height=\"849\" \/><\/p>\n<p>As an intermediate step towards a new generation of structures, the V3.2-Exp introduced DeepSeek Sparse Attention (Note: A Rare Focusing Mechanism) on the basis of V3.1-Terminus, to explore optimization and validation of the training and reasoning efficiency of long texts\u3002<\/p>\n<p>DeepSeek Sparse Attention (DSA) achieved, for the first time, a thin focus mechanism for fine grains, and a significant increase in long text training and reasoning efficiency, with little impact on model output\u3002<\/p>\n<p>In order to critically assess the impact of the introduction of scarce attention, the training set for DeepSeek-V3.2-Exp has been closely aligned with V3.1-Terminus. The performance of DeepSeek-V3.2-Exp in public evaluations in various fields is essentially the same as V3.1-Terminus\u3002<\/p>\n<p>At present, official Apps, webends, and applets are being updated simultaneously to DeepSeek-V3.2-Exp\u3002<\/p>\n<p><strong>This update is here <a href=\"https:\/\/www.1ai.net\/en\/tag\/api\" title=\"_OTHER ORGANISER\" target=\"_blank\" >API<\/a> Large price reductions will reduce the developer 's cost of calling DeepSeek API by 50% or more<\/strong>.<\/p>\n<p>While the validity of DeepSeek-V3.2-Exp has been validated on the public assessment set, it still requires a wider and larger-scale test in the real-life scene of users. Additional API access interfaces have been temporarily reserved for V3.1-Terminus for user-friendly comparison tests\u3002<\/p>\n<p>DeepSeek-V3.2-Exp Model is now in Huggingface with magic<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>:<\/p>\n<ul>\n<li><strong>HuggingFace<\/strong>:<\/li>\n<\/ul>\n<p>https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-V3.2-Exp<\/p>\n<ul>\n<li><strong>ModelScope<\/strong>:<\/li>\n<\/ul>\n<p>https:\/\/modelscope.cn\/models\/deepseek-ai\/DeepSeek-V3.2-Exp<\/p>\n<p>The papers have been published simultaneously:<\/p>\n<p>https:\/\/github.com\/deepseek-ai\/DeepSeek-V3.2-Exp\/blob\/main\/DeepSeek_V3_2.pdf<\/p>","protected":false},"excerpt":{"rendered":"<p>On September 29th, DeepSeek officially released today the DeepSeek-V3.2-Exp model, an experimental version. As an intermediate step towards a new generation of structures, the V3.2-Exp introduced DeepSeek Sparse Attention (Note: A Rare Focusing Mechanism) on the basis of V3.1-Terminus, to explore optimization and validation of the training and reasoning efficiency of long texts. DeepSeek Sparse Attention (DSA) achieved for the first time a thin focus mechanism for fine grains, with long text training and with little impact on model output<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[1033,3606,219],"collection":[],"class_list":["post-44203","post","type-post","status-publish","format-standard","hentry","category-news","tag-api","tag-deepseek","tag-219"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/44203","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=44203"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/44203\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=44203"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=44203"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=44203"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=44203"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}