{"id":55866,"date":"2026-08-15T19:09:33","date_gmt":"2026-08-15T11:09:33","guid":{"rendered":"https:\/\/www.1ai.net\/?p=55866"},"modified":"2026-08-15T19:09:33","modified_gmt":"2026-08-15T11:09:33","slug":"%e6%99%ba%e8%b0%b1%e5%8f%91%e5%b8%83-glm-5-3%ef%bc%8c%e5%90%8c%e4%b8%80%e5%9f%ba%e5%ba%a7%e9%9d%a0%e5%90%8e%e8%ae%ad%e7%bb%83%e6%8f%90%e5%8d%87%e7%bc%96%e7%a8%8b%e4%b8%8e%e7%bd%91%e7%bb%9c%e5%ae%89","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/55866.html","title":{"rendered":"GLM-5.3, BACK-TO-BACK TRAINING ON THE SAME BASE TO ENHANCE PROGRAMMING AND NETWORK SECURITY"},"content":{"rendered":"<p>8 \u6708 15 \u65e5\u6d88\u606f\uff0c<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e6%99%ba%e8%b0%b1\" title=\"\u3010\u67e5\u770b\u542b\u6709[\u667a\u8c31]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">\u667a\u8c31<\/a>\u6628\u65e5\u53d1\u5e03 GLM-5.3\u3002<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e6%96%b0%e6%a8%a1%e5%9e%8b\" title=\"\u3010\u67e5\u770b\u542b\u6709[\u65b0\u6a21\u578b]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">\u65b0\u6a21\u578b<\/a>\u6cbf\u7528 GLM-5.2 \u7684\u57fa\u5ea7\uff0c\u6ca1\u6709\u91cd\u65b0\u9884\u8bad\u7ec3\uff0c\u80fd\u529b\u63d0\u5347\u4e3b\u8981\u6765\u81ea\u66f4\u957f\u65f6\u95f4\u3001\u66f4\u5927\u89c4\u6a21\u548c\u66f4\u591a\u7c7b\u578b\u73af\u5883\u7684\u540e\u8bad\u7ec3\u3002\u5b98\u65b9\u79f0\uff0cGLM-5.3 \u5df2\u4e0a\u7ebf ZCode\u3001AutoClaw\u3001GLM Coding Plan \u4e0e\u591a\u5bb6\u7f16\u7801\u5e73\u53f0\uff0cAPI \u5c06\u968f\u540e\u5f00\u653e\uff0c\u6a21\u578b\u6743\u91cd\u8ba1\u5212\u5728\u5b8c\u6210\u5b89\u5168\u8bc4\u4f30\u548c\u52a0\u56fa\u540e\u4e8e\u4e24\u5468\u5185\u5f00\u6e90\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-55867\" title=\"f3b03cb6j00tjt5mt001nd000u000hqm\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2026\/08\/f3b03cb6j00tjt5mt001nd000u000hqm.jpg\" alt=\"f3b03cb6j00tjt5mt001nd000u000hqm\" width=\"1080\" height=\"638\" \/><\/p>\n<p>\u7f16\u7a0b\uff1aTerminal-Bench 3.0 \u5f97\u5206\u7531 GLM-5.2 \u7684 4.6 \u63d0\u5347\u81f3 28.3\uff0cDeepSWE v1.1 \u7531 46.2 \u63d0\u5347\u81f3 66.9\uff0cAgents&#8217; Last Exam \u7531 23.8 \u63d0\u5347\u81f3 28.5\u3002<\/p>\n<p>\u4ee3\u7801\u6267\u884c\u6548\u7387\uff1a\u667a\u8c31\u81ea\u5efa Z.ai Code Bench \u663e\u793a\uff0cGLM-5.3 \u5728 High \u6863\u4f4d\u51c6\u786e\u7387\u4e3a 31.4%\uff0c\u6bcf\u9879\u4efb\u52a1\u5e73\u5747\u8f93\u51fa\u7ea6 5 \u4e07 token\uff1b\u4f5c\u4e3a\u5bf9\u7167\uff0cClaude Opus 4.8 \u6700\u9ad8\u6863\u4f4d\u4e3a 29.5%\uff0c\u5e73\u5747\u8f93\u51fa\u7ea6 12 \u4e07 token\u3002<\/p>\n<p>\u7f51\u7edc\u5b89\u5168\uff1aGLM-5.3 \u5728 CyberGym \u5f97\u5206 84.5%\uff0c\u7565\u9ad8\u4e8e Mythos 5 \u7684 83.8% \u548c GPT-5.6 Sol \u7684 83.6%\uff1bExploitBench \u5f97\u5206 54.4%\uff0c\u4ecd\u4f4e\u4e8e\u4e24\u6b3e\u95ed\u6e90\u6a21\u578b\u3002<\/p>\n<p>\u5b89\u5168\u8bc4\u4f30\uff1a\u667a\u8c31\u4e0e\u9ad8\u6821\u53ca\u5b89\u5168\u56e2\u961f\u7d2f\u8ba1\u53d1\u73b0\u5e76\u53bb\u91cd 2436 \u4e2a\u6f0f\u6d1e\uff0c\u5176\u4e2d 1097 \u4e2a\u4e3a\u4e2d\u9ad8\u5371\uff0c\u6d89\u53ca 269 \u4e2a\u9879\u76ee\uff1b\u5df2\u4fee\u590d\u6f0f\u6d1e\u4f1a\u8fdb\u5165\u516c\u5f00\u62ab\u9732\u8d26\u672c\uff0c\u4ecd\u5728\u534f\u8c03\u62ab\u9732\u7684\u9879\u76ee\u53ea\u5c55\u793a\u54c8\u5e0c\u503c\u3002<\/p>\n<p>\u667a\u8c31\u540c\u65f6\u542f\u52a8\u300c\u5f00\u6e90\u7684\u76fe\u300d\u8ba1\u5212\uff0c\u5411\u5f00\u6e90\u9879\u76ee\u63d0\u4f9b\u6301\u7eed\u5b89\u5168\u5ba1\u8ba1\u3001\u6a21\u578b\u989d\u5ea6\u4e0e ZCode \u4ee3\u7801\u5ba1\u8ba1\u5165\u53e3\u3002\u516c\u53f8\u79f0\uff0c\u6743\u91cd\u5ef6\u540e\u4e24\u5468\u5f00\u653e\uff0c\u662f\u4e3a\u4e86\u5148\u9650\u5236\u6f5c\u5728\u653b\u51fb\u80fd\u529b\u5e76\u4fdd\u7559\u9632\u5fa1\u7528\u9014\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Message of August 15, the think tank was released yesterday, GM-5.3. The new model follows the GLM-5.2 base, is not retrained, and capacity enhancement comes mainly from post-training in longer, larger and more diverse environments. Officially, GLM-5.3 is online with Zcode, AutoClaw, GLM Coding Plan and multiple coding platforms, and API will then be open, and model weights are scheduled to open within two weeks of the completion of security assessments and enhancements. Programming: Terminal-Bench 3.0 Scores from GM-5.2 4.6 to 28.3, DeepSWE v1.1 from 46.2 to 66.9, Agents &amp; #821<\/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":[1757,2680],"collection":[],"class_list":["post-55866","post","type-post","status-publish","format-standard","hentry","category-news","tag-1757","tag-2680"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/55866","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=55866"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/55866\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=55866"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=55866"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=55866"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=55866"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}