{"id":351,"date":"2025-04-29T02:35:56","date_gmt":"2025-04-29T02:35:56","guid":{"rendered":"https:\/\/minitoolai.com\/blog\/?p=351"},"modified":"2025-04-29T02:35:59","modified_gmt":"2025-04-29T02:35:59","slug":"deepseek-r2-a-new-ai-breakthrough-promising-97-lower-costs-than-gpt-4","status":"publish","type":"post","link":"https:\/\/minitoolai.com\/blog\/deepseek-r2-a-new-ai-breakthrough-promising-97-lower-costs-than-gpt-4\/","title":{"rendered":"DeepSeek R2: A New AI Breakthrough Promising 97% Lower Costs Than GPT-4"},"content":{"rendered":"\n<p>Buzz is building around <a href=\"https:\/\/minitoolai.com\/blog\/what-is-deepseek-the-ultimate-guide-to-using-deepseek-ai\/\" data-type=\"post\" data-id=\"148\">DeepSeek<\/a>&#8216;s upcoming AI model, DeepSeek R2, which is rumored to deliver a major leap forward in artificial intelligence\u2014at a fraction of the cost of competitors like GPT-4. According to circulating reports, the model was trained entirely using Huawei\u2019s Ascend chips, showcasing an impressive move toward self-reliant AI development.<\/p>\n\n\n\n<p>Try DeepSeek for free with unlimited usage: <a href=\"https:\/\/minitoolai.com\/deepseek\/\">minitoolai.com\/deepseek<\/a><\/p>\n\n\n\n<figure class=\"wp-block-image size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"686\" height=\"386\" src=\"https:\/\/minitoolai.com\/blog\/wp-content\/uploads\/2025\/04\/image-94.png\" alt=\"DeepSeek R2: A New AI Breakthrough Promising 97% Lower Costs Than GPT-4\" class=\"wp-image-352\" style=\"width:829px;height:auto\" srcset=\"https:\/\/minitoolai.com\/blog\/wp-content\/uploads\/2025\/04\/image-94.png 686w, https:\/\/minitoolai.com\/blog\/wp-content\/uploads\/2025\/04\/image-94-300x169.png 300w, https:\/\/minitoolai.com\/blog\/wp-content\/uploads\/2025\/04\/image-94-150x84.png 150w\" sizes=\"auto, (max-width: 686px) 100vw, 686px\" \/><figcaption class=\"wp-element-caption\">DeepSeek R2: A New AI Breakthrough Promising 97% Lower Costs Than GPT-4<\/figcaption><\/figure>\n\n\n\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_82_2 counter-hierarchy ez-toc-counter ez-toc-grey ez-toc-container-direction\">\n<div class=\"ez-toc-title-container\">\n<p class=\"ez-toc-title\" style=\"cursor:inherit\">Contents<\/p>\n<span class=\"ez-toc-title-toggle\"><a href=\"#\" class=\"ez-toc-pull-right ez-toc-btn ez-toc-btn-xs ez-toc-btn-default ez-toc-toggle\" aria-label=\"Toggle Table of Content\"><span class=\"ez-toc-js-icon-con\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">Toggle<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #999;color:#999\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewBox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #999;color:#999\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewBox=\"0 0 24 24\" version=\"1.2\" baseProfile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/span><\/a><\/span><\/div>\n<nav><ul class='ez-toc-list ez-toc-list-level-1 ' ><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/minitoolai.com\/blog\/deepseek-r2-a-new-ai-breakthrough-promising-97-lower-costs-than-gpt-4\/#Is_DeepSeek_R2_About_to_Launch\" >Is DeepSeek R2 About to Launch?<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/minitoolai.com\/blog\/deepseek-r2-a-new-ai-breakthrough-promising-97-lower-costs-than-gpt-4\/#Competitive_Performance_at_Game-Changing_Costs\" >Competitive Performance at Game-Changing Costs<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/minitoolai.com\/blog\/deepseek-r2-a-new-ai-breakthrough-promising-97-lower-costs-than-gpt-4\/#Fully_Powered_by_Huaweis_Hardware\" >Fully Powered by Huawei\u2019s Hardware<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/minitoolai.com\/blog\/deepseek-r2-a-new-ai-breakthrough-promising-97-lower-costs-than-gpt-4\/#A_Potential_Earthquake_for_the_AI_Industry\" >A Potential Earthquake for the AI Industry<\/a><\/li><\/ul><\/nav><\/div>\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Is_DeepSeek_R2_About_to_Launch\"><\/span>Is DeepSeek R2 About to Launch?<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>Following the success of its R1 model, DeepSeek has proven that China is more than capable of competing in the high-end AI race. The debut of DeepSeek R1 not only rattled U.S. stock markets, wiping billions off valuations, but also challenged the notion that cutting-edge AI development must be prohibitively expensive\u2014a belief long propagated by companies like OpenAI. Now, rumors from Chinese sources suggest that DeepSeek is preparing to unleash another disruptive force on the AI world with its upcoming R2 model.<\/p>\n\n\n\n<p>It\u2019s important to note that as of now, DeepSeek has not officially confirmed any specifications related to R2. Nonetheless, reports from outlets like Wccftech indicate that DeepSeek R2 will leverage a hybrid Mixture of Experts (MoE) architecture\u2014a step beyond current MoE designs. This new system could incorporate advanced gating mechanisms or blend MoE structures with dense layers to optimize for high-performance tasks. With this setup, R2 is expected to feature a staggering 1.2 trillion parameters, doubling the size of its predecessor.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-rich is-provider-twitter wp-block-embed-twitter\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"twitter-tweet\" data-width=\"550\" data-dnt=\"true\"><p lang=\"en\" dir=\"ltr\">\ud83d\udea8Viral rumors of DeepSeek R2 leaked!<br><br>\u20141.2T param, 78B active, hybrid MoE<br>\u201497.3% cheaper than GPT 4o ($0.07\/M in, $0.27\/M out)<br>\u20145.2PB training data. 89.7% on C-Eval2.0<br>\u2014Better vision. 92.4% on COCO<br>\u201482% utilization in Huawei Ascend 910B<br><br>Big shift away from US supply chain. <a href=\"https:\/\/t.co\/Jncg0PvEYU\">pic.twitter.com\/Jncg0PvEYU<\/a><\/p>&mdash; Deedy (@deedydas) <a href=\"https:\/\/twitter.com\/deedydas\/status\/1916160465958539480?ref_src=twsrc%5Etfw\">April 26, 2025<\/a><\/blockquote><script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script>\n<\/div><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Competitive_Performance_at_Game-Changing_Costs\"><\/span>Competitive Performance at Game-Changing Costs<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>The real headline-grabber isn&#8217;t just R2\u2019s projected power\u2014it\u2019s the cost efficiency. DeepSeek R2 is rumored to rival heavyweights like OpenAI\u2019s GPT-4 Turbo and Google\u2019s Gemini 2.0 Pro in performance, but at dramatically lower prices. Early reports suggest that R2\u2019s processing costs could be 97.4% cheaper than GPT-4, with prices estimated at just $0.07 per million input tokens and $0.27 per million output tokens. If these numbers hold true, DeepSeek R2 could quickly become the most cost-effective large language model on the market, offering businesses a far more affordable alternative to Western AI providers.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Fully_Powered_by_Huaweis_Hardware\"><\/span>Fully Powered by Huawei\u2019s Hardware<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>Another standout feature is DeepSeek&#8217;s strategic use of Huawei\u2019s Ascend 910B chips, achieving an impressive 82% utilization across clusters capable of 512 PetaFLOPS at FP16 precision. This move underscores DeepSeek\u2019s commitment to vertical integration, relying solely on domestic technology for model training\u2014a significant shift in the AI landscape. The close partnership with Huawei highlights a major investment in internal resources, insulating DeepSeek from potential supply chain vulnerabilities linked to foreign hardware.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"A_Potential_Earthquake_for_the_AI_Industry\"><\/span>A Potential Earthquake for the AI Industry<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>Although these details are still based on leaks and speculation, the emerging picture of DeepSeek R2 paints it as a potential game-changer. If the rumors prove accurate, the launch of DeepSeek R2 could shake up the global AI ecosystem, forcing major Western AI firms to rethink their pricing strategies and innovation roadmaps.<\/p>\n\n\n\n<p>For now, the AI world waits with bated breath for official confirmation\u2014but one thing seems clear: DeepSeek R2 could mark a pivotal moment in the evolution of affordable, high-performance artificial intelligence.<\/p>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Buzz is building around DeepSeek&#8216;s upcoming AI model, DeepSeek R2, which is rumored to deliver a major leap forward in artificial intelligence\u2014at a fraction of the cost of competitors like GPT-4. According to circulating reports, the model was trained entirely using Huawei\u2019s Ascend chips, showcasing an impressive move toward self-reliant AI development. Try DeepSeek for [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":352,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5],"tags":[8,29,28,120],"class_list":{"0":"post-351","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-hi-tech","8":"tag-ai","9":"tag-china","10":"tag-deepseek","11":"tag-r2"},"_links":{"self":[{"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/posts\/351","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/comments?post=351"}],"version-history":[{"count":1,"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/posts\/351\/revisions"}],"predecessor-version":[{"id":353,"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/posts\/351\/revisions\/353"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/media\/352"}],"wp:attachment":[{"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/media?parent=351"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/categories?post=351"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/minitoolai.com\/blog\/wp-json\/wp\/v2\/tags?post=351"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}