{"id":325,"date":"2026-10-24T14:05:37","date_gmt":"2026-10-24T06:05:37","guid":{"rendered":"https:\/\/ai.hfbcw.cn\/?p=325"},"modified":"2026-10-24T14:05:39","modified_gmt":"2026-10-24T06:05:39","slug":"kimi-k31-pro-post-training","status":"publish","type":"post","link":"http:\/\/ai.hfbcw.cn\/index.php\/2026\/10\/24\/kimi-k31-pro-post-training\/","title":{"rendered":"\u6ca1\u6362\u57fa\u5ea7\uff0cKimi K3.1 Pro\u9760\u8865\u8bfe\u51b2\u8fdb\u5168\u7403\u524d\u56db"},"content":{"rendered":"

\"\u6ca1\u6362\u57fa\u5ea7\uff0cKimi<\/p>\n

10\u670824\u65e5\u51cc\u6668\uff0c\u6708\u4e4b\u6697\u9762\u53d1\u5e03\u65b0\u65d7\u8230 Kimi K3.1 Pro\u3002\u6700\u53cd\u5e38\u7684\u5730\u65b9\u5728\u4e8e\uff0c\u8fd9\u4e2a”\u65b0\u6a21\u578b”\u538b\u6839\u6ca1\u6362\u57fa\u5ea7\u2014\u2014\u4eca\u5e747\u6708 K3 \u7684\u8001\u5e95\u5b50\u4e00\u52a8\u6ca1\u52a8\uff0c\u53ea\u9760\u4e00\u8f6e\u5927\u89c4\u6a21\u5f3a\u5316\u5b66\u4e60”\u8865\u8bfe”\uff0c\u5c31\u628a\u771f\u5b9e\u7f16\u7a0b\u6d4b\u8bd5\u7684\u6210\u7ee9\u4ece 12.6% \u7ec3\u5230 65.7%\uff0c\u5728\u7b2c\u4e09\u65b9\u673a\u6784 Artificial Analysis \u7684\u667a\u529b\u6307\u6570\u699c\u4e0a\u62ff\u5230 55 \u5206\uff0c\u5750\u5230\u5168\u7403\u7b2c\u56db\uff0c\u548c\u6392\u7b2c\u4e00\u7684 Claude Opus 5.5 \u53ea\u5dee 3 \u5206\u3002\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0c\u82f1\u7279\u5c14CEO\u9648\u7acb\u6b66\u6b64\u524d\u70b9\u540d\u79f0\u8d5eDeepSeek\u3001Kimi\u3001GLM\u591f\u7528\u53c8\u4fbf\u5b9c<\/a>\u2014\u2014\u5982\u4eca\u770b\u6765\uff0c\u8fd9\u4e2a\u5224\u65ad\u6b63\u5728\u88ab K3.1 Pro \u7684\u5b9e\u9645\u8868\u73b0\u5151\u73b0\u3002<\/p>\n

\u5148\u8bf4\u5b83\u5230\u5e95\u5f3a\u5728\u54ea<\/strong><\/p>\n

\u770b\u7f16\u7a0b\u8fd9\u4e00\u9879\u5c31\u591f\u4e86\u3002Terminal-Bench 4.0 \u6d4b\u7684\u4e0d\u662f\u80cc\u9898\uff0c\u662f\u628a\u6a21\u578b\u5355\u72ec\u6254\u8fdb\u4e00\u4e2a\u771f\u5b9e\u7684\u547d\u4ee4\u884c\u73af\u5883\uff0c\u8ba9\u5b83\u81ea\u5df1\u6572\u547d\u4ee4\u3001\u8bfb\u62a5\u9519\u3001\u6539\u4ee3\u7801\uff0c\u76f4\u5230\u628a\u4efb\u52a1\u5e72\u5b8c\u2014\u2014\u50cf\u628a\u5b9e\u4e60\u751f\u4e00\u4e2a\u4eba\u4e22\u8fdb\u673a\u623f\u503c\u591c\u73ed\u3002K3.1 Pro \u5b98\u65b9\u6210\u7ee9 65.7%\uff0c\u8d85\u8fc7 GPT-6 Astra \u7684 58.2%\uff0c\u79bb Claude Opus 5.5 \u7684 66.4% \u53ea\u5dee 0.7 \u4e2a\u767e\u5206\u70b9\uff1bArtificial Analysis \u81ea\u5df1\u590d\u6d4b\uff0c\u751a\u81f3\u7ed9\u51fa\u4e86 67.2%\u3002\u800c\u4e0a\u4e00\u4ee3 K3 \u5728\u540c\u4e00\u9879\u6d4b\u8bd5\u91cc\u53ea\u6709 12.6%\uff0c\u57fa\u672c\u662f\u966a\u8dd1\u6c34\u5e73\u3002<\/p>\n

\u957f\u7a0b\u4efb\u52a1\u66f4\u5938\u5f20\u3002\u5728\u8003\u5bdf\u957f\u65f6\u95f4\u8f6f\u4ef6\u5de5\u7a0b\u7684 SWE-Marathon \u6d4b\u8bd5\u4e2d\uff0c\u5b83\u4ee5 51.7% \u53cd\u8d85 Claude Opus 5.5 \u7684 46.3%\uff1b\u667a\u80fd\u4f53\u8003\u8bd5<\/a> Agents’ Last Exam \u62ff\u5230 43.2 \u5206\uff0c\u6392\u5728\u5b98\u65b9\u5bf9\u6bd4\u6a21\u578b\u7b2c\u4e00\u3002\u91d1\u878d\u6587\u6863\u5206\u6790 OfficeQA Pro v2 \u62ff\u5230 64.4 \u5206\uff0c\u662f\u76ee\u524d\u7684\u6700\u4f18\u6210\u7ee9\u3002<\/p>\n

2.8 \u4e07\u4ebf\u53c2\u6570\uff0c\u5230\u5e95\u662f\u4e2a\u4ec0\u4e48\u6982\u5ff5<\/strong><\/p>\n

K3.1 Pro \u603b\u53c2\u6570 2.8 \u4e07\u4ebf\uff08\u5b98\u65b9\u53e3\u5f84”3 \u4e07\u4ebf\u7ea7”\uff09\uff0c\u4f46\u5b83\u7528\u7684\u662f\u6df7\u5408\u4e13\u5bb6\u67b6\u6784\uff1a\u597d\u6bd4\u4e00\u5bb6\u8bf7\u4e86 2800 \u540d\u4e13\u5bb6\u7684\u516c\u53f8\uff0c\u6bcf\u6b21\u6765\u6d3b\u53ea\u558a\u5176\u4e2d 104 \u540d\u6700\u5bf9\u53e3\u7684\u4e0a\u73ed\uff0c\u5176\u4f59\u4eba\u5f85\u547d\u3002\u6240\u4ee5\u5b83\u8111\u5b50\u5927\uff0c\u7535\u8d39\u5374\u4e0d\u7528\u6309 2.8 \u4e07\u4ebf\u5168\u989d\u4ea4\uff0c\u6bcf\u6b21\u63a8\u7406\u53ea\u6fc0\u6d3b 1040 \u4ebf\u53c2\u6570\u3002<\/p>\n

100 \u4e07 token \u7684\u4e0a\u4e0b\u6587\uff0c\u6362\u7b97\u6210\u4f7f\u7528\u573a\u666f\uff1a\u4e00\u6b21\u80fd\u541e\u4e0b\u7ea6\u4e00\u767e\u4e07\u5b57\uff0c\u76f8\u5f53\u4e8e\u628a\u300a\u7ea2\u697c\u68a6\u300b\u4ece\u5934\u8bfb\u5230\u5c3e\uff0c\u8fd8\u80fd\u518d\u585e\u8fdb\u4e00\u4e2a\u4e2d\u578b\u9879\u76ee\u7684\u5168\u90e8\u4ee3\u7801\u3002\u5bf9\u7a0b\u5e8f\u5458\u6765\u8bf4\uff0c\u4ee5\u524d\u8981\u628a\u4ee3\u7801\u5207\u788e\u4e86\u5582\uff0c\u73b0\u5728\u6574\u4e2a\u4ed3\u5e93\u4e00\u6b21\u5012\u8fdb\u53bb\u5c31\u884c\u3002<\/p>\n

\u5b83\u8fd8\u987a\u624b\u628a 3D \u521b\u4f5c\u62c9\u4e86\u8d77\u6765\uff1a\u80fd\u81ea\u5df1\u5199 Blender \u548c CAD \u7684\u811a\u672c\u5efa\u6a21\uff0c\u80fd\u8c03\u7528\u865a\u5e7b 5.8\u3001Godot \u8fd9\u4e9b\u6e38\u620f\u5f15\u64ce\u505a\u51fa\u53ef\u8fd0\u884c\u7684 3D \u6e38\u620f\uff0c\u5b98\u65b9\u6f14\u793a\u91cc\u751a\u81f3\u7528\u798f\u5efa\u571f\u697c\u7684\u9020\u578b\u8bbe\u8ba1\u4e86\u4e00\u76cf\u80fd 3D \u6253\u5370\u7684\u53f0\u706f\u3002Kimi \u7f51\u9875\u548c App \u91cc\u540c\u6b65\u4e0a\u7ebf\u4e86 3D Make \u529f\u80fd\uff0c\u666e\u901a\u4eba\u4e5f\u80fd\u76f4\u63a5\u4e0a\u624b\u634f\u6a21\u578b\u3002<\/p>\n

\u771f\u6b63\u7684\u72e0\u6d3b\uff0c\u662f\u6ca1\u91cd\u7ec3<\/strong><\/p>\n

\u8fd9\u6b21\u5347\u7ea7\u6700\u503c\u5f97\u7422\u78e8\u7684\uff0c\u4e0d\u662f\u5206\u6570\uff0c\u662f\u65b9\u6cd5\u3002K3.1 Pro \u6ca1\u82b1\u4e00\u5206\u94b1\u5728\u65b0\u7684\u9884\u8bad\u7ec3\u4e0a\uff0c\u800c\u662f\u628a\u5f3a\u5316\u5b66\u4e60\u6269\u5c55\u5230\u6570\u767e\u4e2a\u4e0b\u6e38\u9886\u57df\uff0c\u8ba9\u6a21\u578b\u5728\u771f\u5b9e\u4efb\u52a1\u91cc\u53cd\u590d\u8bd5\u9519\u3001\u6328\u6253\u3001\u4fee\u6b63\u3002\u4ee5\u524d\u884c\u4e1a\u62fc\u7684\u662f\u8c01\u7684\u53d1\u52a8\u673a\u6392\u91cf\u5927\uff0c\u73b0\u5728\u5f00\u59cb\u62fc\u8c01\u7684\u8c03\u6821\u529f\u592b\u6df1\u2014\u2014\u57fa\u5ea7\u8fd8\u662f\u90a3\u53f0\u57fa\u5ea7\uff0c\u8dd1\u51fa\u6765\u7684\u6210\u7ee9\u5b8c\u5168\u4e24\u4e2a\u6837\u3002<\/p>\n

\u8fde”\u4e00\u672c\u6b63\u7ecf\u80e1\u8bf4\u516b\u9053”\u90fd\u6536\u655b\u4e86\uff1a\u5355\u6761\u9648\u8ff0\u7684\u5e7b\u89c9\u7387\u4ece\u4e0a\u4e00\u4ee3\u7684 23.4% \u964d\u5230 8.3%\u3002\u5bf9\u91d1\u878d\u3001\u6cd5\u52a1\u8fd9\u7c7b\u5bb9\u9519\u7387\u4e3a\u96f6\u7684\u573a\u666f\uff0c\u8fd9\u4e2a\u6570\u5b57\u6bd4\u8dd1\u5206\u91cd\u8981\u5f97\u591a\u3002<\/p>\n

\u4ef7\u683c\uff0c\u624d\u662f\u8bf4\u7ed9\u5f00\u53d1\u8005\u7684\u5b9e\u8bdd<\/strong><\/p>\n

\u53c2\u6570\u548c\u8dd1\u5206\u662f\u9762\u5b50\uff0c\u4ef7\u683c\u662f\u91cc\u5b50\u3002K3.1 Pro \u7684 API \u5b9a\u4ef7\u662f\u6bcf\u767e\u4e07 token \u8f93\u5165 2 \u7f8e\u5143\u3001\u8f93\u51fa 10 \u7f8e\u5143\uff0c\u6bd4 K3 \u6574\u4f53\u964d\u4e86 33%\uff1b\u540c\u6b65\u63a8\u51fa\u7684\u9ad8\u901f\u7248 Highspeed \u672c\u4e8b\u4e00\u6837\uff0c\u5410\u5b57\u901f\u5ea6\u5feb 2 \u5230 3 \u500d\u3002Artificial Analysis \u7b97\u8fc7\u4e00\u7b14\u8d26\uff1a\u5b83\u5b8c\u6210\u5355\u4e2a\u4efb\u52a1\u7684\u5e73\u5747\u82b1\u8d39\u7ea6 1.98 \u7f8e\u5143\uff0c\u548c\u524d\u4ee3\u6301\u5e73\u3002\u6362\u53e5\u8bdd\u8bf4\uff0c14 \u5757\u94b1\u5de6\u53f3\uff0c\u5c31\u80fd\u8ba9 AI \u72ec\u7acb\u5e72\u5b8c\u4e00\u4e2a\u771f\u5b9e\u7684\u7f16\u7a0b\u4efb\u52a1\u2014\u2014\u6bd4\u4f60\u7ed9\u5b9e\u4e60\u751f\u70b9\u7684\u90a3\u676f\u5976\u8336\u8fd8\u4fbf\u5b9c\u3002<\/p>\n

\u5f00\u6e90\u4e4b\u524d\uff0c\u5148\u81ea\u65ad\u4e00\u81c2<\/strong><\/p>\n

\u8fd8\u6709\u4e2a\u5bb9\u6613\u88ab\u5ffd\u7565\u7684\u7ec6\u8282\uff1a\u6708\u4e4b\u6697\u9762\u4e3b\u52a8\u4ece\u8bad\u7ec3\u6570\u636e\u91cc\u5220\u6389\u4e86\u7f51\u7edc\u653b\u9632\u76f8\u5173\u7684\u5185\u5bb9\uff0c\u6a21\u578b\u5728\u6f0f\u6d1e\u5229\u7528\u6d4b\u8bd5\u4e2d\u7684\u901a\u8fc7\u7387\u4ece 90.5% \u538b\u5230 81.0%\uff0c\u8def\u900f\u793e\u4e13\u95e8\u53d1\u4e86\u62a5\u9053\u3002\u80cc\u666f\u662f\u4e2d\u56fd\u521a\u5c31 AI \u6076\u610f\u8f6f\u4ef6\u751f\u6210\u6280\u672f\u63d0\u51fa\u51fa\u53e3\u7ba1\u5236\u8349\u6848\u3002K3.1 Pro \u7684\u5b8c\u6574\u6743\u91cd\u5c06\u5728 11 \u6708 4 \u65e5\u524d\u5f00\u653e\u2014\u2014\u5f00\u6e90\u6743\u91cd\u4e00\u65e6\u653e\u51fa\uff0c\u5c31\u518d\u4e5f\u6536\u4e0d\u56de\u6765\uff0c\u6240\u4ee5\u5148\u628a\u6700\u5371\u9669\u7684\u90a3\u90e8\u5206\u80fd\u529b\u6441\u4e0b\u53bb\u3002\u8fd9\u5927\u6982\u4f1a\u6210\u4e3a\u4ee5\u540e\u6240\u6709\u5f00\u6e90\u5927\u5382\u7684\u51fa\u5382\u52a8\u4f5c\uff1a\u80fd\u529b\u8d8a\u5f3a\uff0c\u653e\u51fa\u6765\u4e4b\u524d\u8d8a\u5f97\u5148\u8bc1\u660e\u81ea\u5df1\u514b\u5236\u3002<\/p>\n

\u6700\u540e\u4e00\u53e5<\/strong><\/p>\n

\u6768\u690d\u9e9f\u5e74\u521d\u653e\u8fc7\u8bdd\uff0c\u8bf4\u8981\u4e00\u8def\u505a\u5230 K4\u3001K5\uff0c\u76f4\u81f3 K100\u3002\u8fd9\u6b21 K3.1 Pro \u5148\u8bc1\u660e\u4e86\u4e00\u4ef6\u4e8b\uff1a\u901a\u5f80 K100 \u7684\u8def\u4e0a\uff0c\u4e0d\u5fc5\u6bcf\u4e00\u6b65\u90fd\u91cd\u7ec3\u57fa\u5ea7\uff0c\u65e7\u529f\u5e95\u9760\u540e\u8bad\u7ec3\u7167\u6837\u80fd\u903c\u51fa\u65b0\u4e0a\u9650\u3002<\/p>\n

\u5f53\u7136\uff0c\u8dd1\u5206\u5927\u591a\u8fd8\u662f\u5b98\u65b9\u53e3\u5f84\u3002\u5f00\u6e90\u6a21\u578b<\/a>\u548c\u95ed\u6e90\u6a21\u578b\u7684\u5206\u5dee\u5df2\u7ecf\u4ece 12 \u5206\u7f29\u5230 3 \u5206\uff0c\u662f\u771f\u662f\u5047\uff0c\u7b49 11 \u6708 4 \u65e5\u6743\u91cd\u843d\u5730\u3001\u5168\u7403\u5f00\u53d1\u8005\u81ea\u5df1\u4e0a\u624b\u8dd1\u4e00\u904d\uff0c\u81ea\u6709\u516c\u8bba\u3002\u5728\u90a3\u4e4b\u524d\uff0c\u8fd9\u4e2a\u4fe1\u53f7\u503c\u5f97\u8bb0\u4f4f\uff1a\u56fd\u4ea7\u6a21\u578b\u8fd9\u4e00\u8f6e\u6da8\u7684\u4e0d\u662f\u53c2\u6570\uff0c\u662f\u5185\u529f\u3002<\/p>\n

\u2014\u2014\u718a\u53d4\u7684\u8305\u8349\u5c4b<\/p>\n

\u6570\u636e\u6765\u6e90\uff1a\u6708\u4e4b\u6697\u9762\u5b98\u65b9\u53d1\u5e03\u3001\u8def\u900f\u793e\u3001\u817e\u8baf\u65b0\u95fb\u3001\u51e4\u51f0\u7f51\u3001\u65b0\u6d6a\u8d22\u7ecf\u3001Artificial Analysis \u516c\u5f00\u6d4b\u8bc4\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

\u6708\u4e4b\u6697\u976210\u670824\u65e5\u53d1\u5e03Kimi K3.1 Pro\uff1a\u6ca1\u6362K3\u57fa\u5ea7\uff0c\u53ea\u9760\u5927\u89c4\u6a21\u5f3a\u5316\u5b66\u4e60\u8865\u8bfe\uff0c\u7f16\u7a0b\u8dd1\u5206\u4ece12.6%\u8dc3\u5347\u81f365.7%\uff0c\u51b2\u8fdb\u5168\u7403\u524d\u56db\u3002\u672c\u6587\u62c6\u89e3\u5b83\u9760\u540e\u8bad\u7ec3\u903c\u51fa\u65e7\u57fa\u5ea7\u4e0a\u9650\u7684\u65b9\u6cd5\u3001\u964d\u4ef733%\u7684API\u8d26\uff0c\u4ee5\u53ca\u5f00\u6e90\u524d\u5148\u81ea\u65ad\u7f51\u7edc\u653b\u9632\u4e00\u81c2\u7684\u514b\u5236\u3002<\/p>\n","protected":false},"author":2,"featured_media":324,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[7],"tags":[],"class_list":["post-325","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-pinglun"],"menu_order":0,"_links":{"self":[{"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/posts\/325","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/comments?post=325"}],"version-history":[{"count":0,"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/posts\/325\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/media\/324"}],"wp:attachment":[{"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/media?parent=325"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/categories?post=325"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/ai.hfbcw.cn\/index.php\/wp-json\/wp\/v2\/tags?post=325"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}