{"id":32448,"date":"2025-04-06T15:41:48","date_gmt":"2025-04-06T07:41:48","guid":{"rendered":"https:\/\/www.1ai.net\/?p=32448"},"modified":"2025-04-06T15:41:48","modified_gmt":"2025-04-06T07:41:48","slug":"meta-%e5%8f%91%e5%b8%83-llama-4-%e7%b3%bb%e5%88%97-ai%e6%a8%a1%e5%9e%8b%ef%bc%8c%e5%bc%95%e5%85%a5%e6%b7%b7%e5%90%88%e4%b8%93%e5%ae%b6%e6%9e%b6%e6%9e%84%e6%8f%90%e5%8d%87%e6%95%88","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/32448.html","title":{"rendered":"Meta Releases Llama 4 Series of AI Models, Introduces \"Hybrid Expert Architecture\" to Improve Efficiency"},"content":{"rendered":"

4 \u6708 6 \u65e5\u6d88\u606f\uff0cMeta<\/a> \u73b0\u5df2\u53d1\u5e03\u65d7\u4e0b\u6700\u65b0 Llama 4<\/a> \u7cfb\u5217 AI\u6a21\u578b<\/a>\uff0c\u5305\u542b Llama 4 Scout\u3001Llama 4 Maverick \u548c Llama 4 Behemoth\uff0cMeta \u900f\u9732\u76f8\u5e94\u6a21\u578b\u90fd\u7ecf\u8fc7\u4e86\u201c\u5927\u91cf\u672a\u6807\u6ce8\u7684\u6587\u672c\u3001\u56fe\u50cf\u548c\u89c6\u9891\u6570\u636e\u201d\u7684\u8bad\u7ec3\uff0c\u4ee5\u4f7f\u5b83\u4eec\u5177\u5907\u201c\u5e7f\u6cdb\u7684\u89c6\u89c9\u7406\u89e3\u80fd\u529b\u201d<\/strong>\u3002<\/p>\n

\"5ce2b7f3j00suadcl00agd000v900hkp\"<\/p>\n

\u76ee\u524d\uff0cMeta \u5df2\u5c06\u7cfb\u5217\u6a21\u578b\u4e2d\u7684 Scout \u548c\u00a0Maverick \u4e0a\u67b6\u5230\u00a0Hugging Face\uff08\u70b9\u6b64\u8bbf\u95ee\uff09\uff0c\u800c Behemoth \u6a21\u578b\u5219\u8fd8\u5728\u8bad\u7ec3\u4e2d\u3002\u5176\u4e2d Scout \u53ef\u4ee5\u5728\u5355\u4e2a\u82f1\u4f1f\u8fbe\u00a0H100 GPU \u4e0a\u8fd0\u884c\uff0c\u800c\u00a0Maverick \u5219\u9700\u8981\u82f1\u4f1f\u8fbe\u00a0H100 DGX AI \u5e73\u53f0\u6216\u201c\u540c\u7b49\u6027\u80fd\u7684\u8bbe\u5907\u201d\u3002<\/p>\n

Meta \u8868\u793a\uff0cLlama 4 \u7cfb\u5217\u6a21\u578b\u662f\u8be5\u516c\u53f8\u65d7\u4e0b\u9996\u6279\u91c7\u7528\u6df7\u5408\u4e13\u5bb6\uff08MoE\uff09\u67b6\u6784\u7684\u6a21\u578b<\/strong>\uff0c\u8fd9\u79cd\u67b6\u6784\u5728\u8bad\u7ec3\u548c\u56de\u7b54\u7528\u6237\u67e5\u8be2\u65f6\u7684\u6548\u7387\u66f4\u9ad8\uff0c\u6240\u8c13\u201c\u6df7\u5408\u4e13\u5bb6\u67b6\u6784\u201d\u57fa\u672c\u4e0a\u662f\u5c06\u6570\u636e\u5904\u7406\u4efb\u52a1\u5206\u89e3\u4e3a\u5b50\u4efb\u52a1\uff0c\u7136\u540e\u5c06\u5b83\u4eec\u59d4\u6d3e\u7ed9\u66f4\u5c0f\u7684\u3001\u4e13\u95e8\u7684\u201c\u4e13\u5bb6\u201d\u6a21\u578b\u3002<\/p>\n

\u4f8b\u5982\uff0cMaverick \u603b\u5171\u6709\u00a04000 \u4ebf\u4e2a\u53c2\u6570<\/strong>\uff0c\u4f46\u5728\u00a0128 \u4e2a\u201c\u4e13\u5bb6\u201d\u6a21\u578b\u4e2d\u53ea\u6709\u00a0170 \u4ebf\u4e2a\u6d3b\u8dc3\u53c2\u6570\uff08\u53c2\u6570\u6570\u91cf\u5927\u81f4\u4e0e\u6a21\u578b\u7684\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u76f8\u5bf9\u5e94\uff09\u3002Scout \u6709\u00a0170 \u4ebf\u4e2a\u6d3b\u8dc3\u53c2\u6570<\/strong>\u300116 \u4e2a\u201c\u4e13\u5bb6\u201d\u6a21\u578b\u4ee5\u53ca\u603b\u8ba1\u00a01090 \u4ebf\u4e2a\u53c2\u6570\u3002<\/p>\n

\u4e0d\u8fc7\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cLlama 4 \u7cfb\u5217\u4e2d\u6ca1\u6709\u4e00\u4e2a\u6a21\u578b\u662f\u50cf\u00a0OpenAI \u7684\u00a0o1 \u548c\u00a0o3-mini \u90a3\u6837\u771f\u6b63\u610f\u4e49\u4e0a\u7684\u201c\u63a8\u7406\u6a21\u578b\u201d\u3002\u4f5c\u4e3a\u6bd4\u8f83\uff0c\u201c\u63a8\u7406\u6a21\u578b\u201d\u4f1a\u5bf9\u5176\u7b54\u6848\u8fdb\u884c\u4e8b\u5b9e\u6838\u67e5\uff0c\u5e76\u4e14\u901a\u5e38\u80fd\u66f4\u53ef\u9760\u5730\u56de\u7b54\u95ee\u9898\uff0c\u4f46\u56e0\u6b64\u4e5f\u6bd4\u4f20\u7edf\u7684\u201c\u975e\u63a8\u7406\u201d\u6a21\u578b\u82b1\u8d39\u66f4\u957f\u7684\u65f6\u95f4\u6765\u7ed9\u51fa\u7b54\u6848\u3002<\/p>\n

Meta \u5185\u90e8\u6d4b\u8bd5\u663e\u793a\uff0cMaverick \u6a21\u578b\u6700\u9002\u5408\u7528\u4e8e\u201c\u901a\u7528\u00a0AI \u52a9\u624b\u548c\u804a\u5929\u201d\u7b49\u5e94\u7528\u573a\u666f\uff0c\u8be5\u6a21\u578b\u5728\u521b\u610f\u5199\u4f5c\u3001\u4ee3\u7801\u751f\u6210\u3001\u7ffb\u8bd1\u3001\u63a8\u7406\u3001\u957f\u6587\u672c\u4e0a\u4e0b\u6587\u603b\u7ed3\u548c\u56fe\u50cf\u57fa\u51c6\u6d4b\u8bd5\u4e2d\u8868\u73b0\u8d85\u8fc7\u4e86\u00a0OpenAI \u7684\u00a0GPT-4o \u548c\u8c37\u6b4c\u7684\u00a0Gemini 2.0 \u7b49\u6a21\u578b\u3002\u7136\u800c Maverick \u4e0e\u8c37\u6b4c\u7684 Gemini 2.5 Pro\u3001Anthropic \u7684\u00a0Claude 3.7 Sonnet \u548c\u00a0OpenAI \u7684\u00a0GPT-4.5 \u7b49\u529f\u80fd\u66f4\u5f3a\u5927\u7684\u6700\u65b0\u6a21\u578b\u76f8\u6bd4\u4ecd\u6709\u4e00\u5b9a\u8fdb\u6b65\u7a7a\u95f4\u3002<\/p>\n

\u800c Scout \u7684\u4f18\u52bf\u5219\u662f\u603b\u7ed3\u6587\u6863\u3001\u57fa\u4e8e\u5927\u578b\u4ee3\u7801\u5e93\u8fdb\u884c\u63a8\u7406\u3002\u8be5\u6a21\u578b\u652f\u6301 1000 \u4e07\u4e2a\u8bcd\u5143\uff08\u201c\u8bcd\u5143\u201d\u4ee3\u8868\u539f\u59cb\u6587\u672c\u7684\u7247\u6bb5\uff0c\u4f8b\u5982\u5355\u8bcd\u201cfantastic\u201d\u53ef\u62c6\u5206\u4e3a\u201cfan\u201d\u3001\u201ctas\u201d\u548c\u201ctic\u201d\uff09\uff0c\u56e0\u6b64\u4e00\u6b21\u53ef\u4ee5\u5904\u7406\u201c\u591a\u8fbe\u6570\u767e\u4e07\u5b57\u7684\u6587\u672c\u201d\u3002<\/p>\n

1AI\u6ce8\u610f\u5230\uff0cMeta \u8fd8\u8fdb\u4e00\u6b65\u9884\u544a\u4e86\u5176 Behemoth \u6a21\u578b\uff0c\u636e\u8be5\u516c\u53f8\u79f0\uff0cBehemoth \u6709\u00a02880 \u4ebf\u4e2a\u6d3b\u8dc3\u53c2\u6570<\/strong>\u300116 \u4e2a\u201c\u4e13\u5bb6\u201d\u6a21\u578b\uff0c\u603b\u53c2\u6570\u6570\u91cf\u63a5\u8fd1\u00a02 \u4e07\u4ebf\u4e2a\u3002Meta \u7684\u5185\u90e8\u57fa\u51c6\u6d4b\u8bd5\u663e\u793a\uff0c\u5728\u4e00\u4e9b\u8861\u91cf\u89e3\u51b3\u6570\u5b66\u95ee\u9898\u7b49\u79d1\u5b66\u3001\u6280\u672f\u3001\u5de5\u7a0b\u548c\u6570\u5b66\uff08STEM\uff09\u6280\u80fd\u7684\u8bc4\u4f30\u4e2d\uff0cBehemoth \u7684\u8868\u73b0\u4f18\u4e8e\u00a0GPT-4.5\u3001Claude 3.7 Sonnet \u548c\u00a0Gemini 2.0 Pro\uff0c\u4f46\u4e0d\u5982\u00a0Gemini 2.5 Pro\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

On April 6, Meta has released the latest Llama 4 series of AI models under its flag, which include Llama 4 Scott, Llama 4 Maverick and Llama 4 Behemoth, and Meta has revealed that the corresponding models have been trained in \u201cmany unmarked text, images and video data\u201d to provide them with \u201ca wide range of visual understanding\u201d. At present, Meta has mounted Scout and Maverick in a series of models to Hugging Face, while the Behemoth model is still in training. Of which scout can run on a single engweeda H100 GPU, while M<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[146],"tags":[167,184,6188,297],"collection":[],"class_list":["post-32448","post","type-post","status-publish","format-standard","hentry","category-news","tag-ai","tag-llama","tag-llama-4","tag-meta"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/32448","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=32448"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/32448\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=32448"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=32448"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=32448"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=32448"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}