{"id":59923,"date":"2026-11-10T09:33:20","date_gmt":"2026-11-10T01:33:20","guid":{"rendered":"https:\/\/www.1ai.net\/?p=59923"},"modified":"2026-11-10T09:33:20","modified_gmt":"2026-11-10T01:33:20","slug":"meta-%e6%8e%a8%e5%87%ba%e8%bd%bb%e9%87%8f%e7%ba%a7-ai%e6%a8%a1%e5%9e%8b-muse-ember%ef%bc%9a%e8%b7%91%e5%88%86%e5%aa%b2%e7%be%8e-claude-haiku-5-5%ef%bc%8c%e5%8d%95%e4%bb%bb%e5%8a%a1%e6%88%90%e6%9c%ac","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/59923.html","title":{"rendered":"Meta rolls out lightweight AI model Muse Ember: Runs a fair run Claude Haiku 5.5, single job cost only 1\/3"},"content":{"rendered":"

11 \u6708 10 \u65e5\u6d88\u606f\uff0cMeta<\/a> \u73b0\u5df2\u63a8\u51fa Muse Ember AI\u6a21\u578b<\/a>\u3002\u8fd9\u662f Meta \u8fc4\u4eca\u6700\u5c0f\u3001\u6548\u7387\u6700\u9ad8\u7684\u6a21\u578b\uff0c\u4e13\u4e3a\u590d\u6742\u63a8\u7406\u4efb\u52a1\u6267\u884c\u4ee5\u53ca\u667a\u80fd\u4f53\u65e5\u5e38\u53cd\u590d\u5904\u7406\u7684\u5de5\u4f5c\u800c\u8bbe\u8ba1\u3002<\/p>\n

Meta \u8868\u793a\uff0c\u5728 Artificial Analysis \u6392\u884c\u699c\u4e2d\uff0c\u8fd9\u6b3e\u6a21\u578b\u7684\u8868\u73b0\u53ef\u4ee5\u5ab2\u7f8e Claude Haiku 5.5<\/strong>\uff0c\u4f46\u5355\u4efb\u52a1\u6210\u672c\u4ec5\u4e3a 1\/3<\/strong>\uff0c\u5e76\u5728\u591a\u9879\u6d4b\u8bd5\u4e2d\u8d85\u8d8a GPT-6 Luna\u3002<\/p>\n

\"8f984d3bj00to9iwm004md0014000n6m\"<\/p>\n

1AI \u9644 Muse Ember \u6a21\u578b\u8dd1\u5206\u6570\u636e\u5982\u4e0b\uff1a<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
<\/th>\n\n

\u8bc4\u6d4b\u9879\u76ee<\/p>\n<\/th>\n

\n

Muse Ember\uff08max\uff09<\/p>\n<\/th>\n

\n

Claude Haiku 5.5\uff08max\uff09<\/p>\n<\/th>\n

\n

GPT-6 Luna\uff08max\uff09<\/p>\n<\/th>\n<\/tr>\n

\u667a\u80fd\u4f53<\/td>\n\n

AutomationBench-AA\uff08\u7aef\u5230\u7aef\u4e1a\u52a1\u5de5\u4f5c\u6d41\uff09<\/p>\n<\/td>\n

\n

37.8<\/p>\n<\/td>\n

\n

13.9<\/p>\n<\/td>\n

\n

20.2<\/p>\n<\/td>\n<\/tr>\n

\n

Agents’ Last Exam\uff08\u7aef\u5230\u7aef\u667a\u80fd\u4f53\u4efb\u52a1\uff09<\/p>\n<\/td>\n

\n

61.6<\/p>\n<\/td>\n

\n

57.0<\/p>\n<\/td>\n

\n

57.1<\/p>\n<\/td>\n<\/tr>\n

\n

Personal Superintelligence Index\uff08\u4e2a\u4eba\u4e0e\u77e5\u8bc6\u5de5\u4f5c\u5185\u90e8\u7efc\u5408\u6307\u6570\uff09<\/p>\n<\/td>\n

\n

46.8<\/p>\n<\/td>\n

\n

36.1<\/p>\n<\/td>\n

\n

37.0<\/p>\n<\/td>\n<\/tr>\n

\u7f16\u7a0b<\/td>\n\n

Terminal-Bench 4.0\uff08\u667a\u80fd\u4f53\u7ec8\u7aef\u7f16\u7a0b\uff09<\/p>\n<\/td>\n

\n

40.9<\/p>\n<\/td>\n

\n

39.2<\/p>\n<\/td>\n

\n

16.4<\/p>\n<\/td>\n<\/tr>\n

\n

Coding Agent Index-AA\uff08\u771f\u5b9e\u667a\u80fd\u4f53\u73af\u5883\u4e2d\u7684\u7f16\u7a0b\uff09<\/p>\n<\/td>\n

\n

46<\/p>\n<\/td>\n

\n

37<\/p>\n<\/td>\n

\n

31<\/p>\n<\/td>\n<\/tr>\n

\n

FrontierCode 2.0-AA\uff08\u4ee5\u4ee3\u7801\u5408\u5e76\u53ef\u884c\u6027\u4e3a\u6807\u51c6\u8bc4\u4f30\u7684\u771f\u5b9e\u5de5\u7a0b\u4efb\u52a1\uff09<\/p>\n<\/td>\n

\n

27.2<\/p>\n<\/td>\n

\n

34.9<\/p>\n<\/td>\n

\n

25.2<\/p>\n<\/td>\n<\/tr>\n

\u591a\u6a21\u6001<\/td>\n\n

GDP.pdf-AA<\/span>\uff08\u957f\u6587\u6863\u4e13\u4e1a\u5de5\u4f5c\uff09<\/p>\n<\/td>\n

\n

22.8<\/p>\n<\/td>\n

\n

20.8<\/p>\n<\/td>\n

\n

22.8<\/p>\n<\/td>\n<\/tr>\n

\n

OSWorld 2.1\uff08\u8de8\u684c\u9762\u5e94\u7528\u7684\u8ba1\u7b97\u673a\u64cd\u4f5c\uff09<\/p>\n<\/td>\n

\n

69.5<\/p>\n<\/td>\n

\n

68.9<\/p>\n<\/td>\n

\n

50.5<\/p>\n<\/td>\n<\/tr>\n

\n

PerceptionBench\uff08\u7cbe\u7ec6\u89c6\u89c9\u611f\u77e5\uff09<\/p>\n<\/td>\n

\n

80.6<\/p>\n<\/td>\n

\n

78.6<\/p>\n<\/td>\n

\n

62.8<\/p>\n<\/td>\n<\/tr>\n

\u7efc\u5408<\/td>\n\n

Artificial Analysis Index<\/p>\n<\/td>\n

\n

36<\/p>\n<\/td>\n

\n

36<\/p>\n<\/td>\n

\n

28<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

\u540c\u65f6\uff0c\u8fd9\u6b3e\u6a21\u578b\u62e5\u6709 100 \u4e07 Token \u4e0a\u4e0b\u6587\u7a97\u53e3<\/strong>\uff0c\u6700\u591a\u652f\u6301\u8f93\u51fa 51.2 \u4e07 Token\uff0c\u53ef\u4ee5\u5728\u4e00\u6b21\u8c03\u7528\u4e2d\u5904\u7406\u5b8c\u6574\u7684\u4ee3\u7801\u5e93\u6216\u4e00\u6574\u5957\u6587\u6863\u3002\u5b83\u53ef\u4ee5\u8ba9\u667a\u80fd\u4f53\u5728\u4e00\u6b21\u4f1a\u8bdd\u4e2d\u6267\u884c\u6570\u767e\u6b21\u7f16\u7a0b\u5b50\u4efb\u52a1\uff0c\u5305\u542b\u8bfb\u53d6\u6587\u4ef6\u3001\u8fd0\u884c\u6d4b\u8bd5\u548c\u8fdb\u884c\u6709\u9488\u5bf9\u6027\u7684\u4fee\u590d\u3002<\/p>\n

Muse Ember \u539f\u751f\u652f\u6301\u7406\u89e3\u6587\u672c\u3001\u56fe\u50cf\u3001\u97f3\u9891\u548c\u89c6\u9891\uff0c\u667a\u80fd\u4f53\u65e0\u9700\u501f\u52a9\u5176\u4ed6\u6a21\u578b\uff0c\u5c31\u80fd\u8bfb\u53d6\u622a\u56fe\u3001\u56fe\u8868\u548c\u626b\u63cf\u6587\u6863\uff0c\u4e5f\u80fd\u5904\u7406\u8bed\u97f3\u5907\u5fd8\u5f55\u3001\u4f1a\u8bae\u5f55\u97f3\u548c\u89c6\u9891\u3002\u5b83\u80fd\u5904\u7406\u7bc7\u5e45\u8f83\u957f\u7684\u771f\u5b9e\u6587\u6863\uff0c\u76f4\u63a5\u6839\u636e\u5c4f\u5e55\u622a\u56fe\u64cd\u4f5c\u8ba1\u7b97\u673a\uff0c\u5e76\u8bc6\u522b\u7ec6\u5fae\u7684\u89c6\u89c9\u7ec6\u8282\u3002<\/p>\n

\u5b89\u5168\u65b9\u9762\uff0c\u8fd9\u6b3e\u6a21\u578b\u5728\u53d1\u5e03\u524d\u7ecf\u8fc7 Meta SI \u6269\u5c55\u6846\u67b6\u8bc4\u4f30\uff0c\u6240\u6709\u6d4b\u8bd5\u7ed3\u679c\u663e\u793a\uff0cMuse Ember \u7684\u98ce\u9669\u6c34\u5e73\u4f4e\u4e8e\u6846\u67b6\u8bbe\u5b9a\u7684\u9ad8\u98ce\u9669\u9608\u503c\u3002<\/p>\n

\u4ef7\u683c\u65b9\u9762\uff0c\u8fd9\u6b3e\u6a21\u578b\u5373\u65e5\u8d77\u5728 Meta Model API\u3001Google Cloud \u548c Oracle Cloud \u63d0\u4f9b\u3002\u8f93\u5165\u4ef7\u683c\u4e3a\u6bcf\u767e\u4e07 Token 0.10 \u7f8e\u5143<\/strong>\uff08IT\u4e4b\u5bb6\u6ce8\uff1a\u73b0\u6c47\u7387\u7ea6\u5408 0.67 \u5143\u4eba\u6c11\u5e01\uff09<\/strong>\uff0c\u8f93\u51fa\u4ef7\u683c\u4e3a\u6bcf\u767e\u4e07 Token 0.50 \u7f8e\u5143\uff08\u73b0\u6c47\u7387\u7ea6\u5408 3.3 \u5143\u4eba\u6c11\u5e01\uff09\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

On November 10th, Meta has now launched the Muse Ember AI model. This is the smallest and most efficient model to date, designed for complex reasoning missions and for routine and iterative work by intelligent bodies. Meta states that in the Artificial Analysis ranking, the model performed as much as Claude Haiku 5.5, but the single task cost was only 1\/3 and exceeded GPT-6 Luna in several tests. 1AI with Muse Ember model run data as follows:<\/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":[167,297],"collection":[],"class_list":["post-59923","post","type-post","status-publish","format-standard","hentry","category-news","tag-ai","tag-meta"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/59923","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=59923"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/59923\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=59923"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=59923"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=59923"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=59923"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}