[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fI4fMsJuqdwzJdzhI4sJXRDy7bEntsqhueR1dUY5fHhw":3,"$fWNsfe5LyMk_QM7fXcoufrWVu1aa3Gn4kGdkU0eJxsuI":81},{"id":4,"name":5,"code":6,"version":7,"logo":8,"url":9,"vendorId":10,"vendorName":11,"vendorCode":12,"vendorCountry":13,"modelType":18,"openSource":18,"openLicense":35,"paramSize":36,"contextLength":37,"intro":38,"supportApi":18,"apiUrl":39,"pricingType":40,"freeQuota":41,"score":28,"isRecommend":18,"isHot":18,"click":42,"tags":43,"content":44,"pricings":45,"scores":54},48,"智谱 GLM-5.3","glm-5.3","5.3","\u002Flogos\u002Fzhipu.svg",null,12,"智谱AI","zhipu",[14,19,23,27,31],{"id":15,"name":16,"color":17,"type":18},4,"开源","#909399",1,{"id":20,"name":21,"color":22,"type":18},5,"支持API","#00B4A0",{"id":24,"name":25,"color":26,"type":18},8,"国产自研","#F56C6C",{"id":28,"name":29,"color":30,"type":18},9,"推理强化","#E65100",{"id":32,"name":33,"color":34,"type":18},10,"长上下文","#3B55D9","GLM-5.3 License","753B(MoE)",1000000,"2026 年 8 月发布的国产开源旗舰，编程与网络安全能力突出，已开源","https:\u002F\u002Fopen.bigmodel.cn\u002Fdev\u002Fapi",2,"详见官网当期免费额度活动",17,[],"GLM-5.3 于 2026 年 8 月 14 日发布（权重于两周后开源），与 GLM-5.2 共享同一底座，全部提升来自后训练 Scaling 的优化——可以理解为课本没变、训练方法换了。总参数规模约 7530 亿（MoE 架构，公开报道中亦有 7430 亿口径），上下文长度 100 万 token，最大输出 12.8 万 token。官方称内部体感评测中编程能力较 GLM-5.2 提升 50%，在 CyberGym 漏洞推理评测中得 84.5%，高于 Mythos 5 的 83.8% 与 GPT-5.6 Sol 的 83.6%；更值得注意的是它在 269 个真实开源项目中累计发现 2436 个漏洞（其中 1097 个中高危），这是国产模型少见的「涌现型安全能力」。推理强制开启，可用 reasoning_effort 在 low \u002F high \u002F max 之间权衡成本与深度。许可证为自定义的 GLM-5.3 License：一般用户可自由使用、修改与分发，仅对连续 12 个月合计营收超过 100 亿美元且从事 MaaS 的企业要求先通过 Z.ai 安全审查。同系列另有 GLM-5.3-Flash（8 月 26 日开源，定价约为旗舰版十分之一）与主打开源代码能力的版本，本文一并归入本条目。",[46],{"itemName":47,"billingType":18,"inputPrice":48,"outputPrice":49,"price":50,"unit":51,"currency":52,"freeQuota":9,"intro":53},"标准 API 调用",1.4,4.4,0,"美元\u002F百万tokens","USD","GLM-5.3-Flash 约为旗舰版十分之一",[55,58,62,66,70,74,78],{"metricId":18,"metricName":56,"metricCode":57,"maxScore":32,"score":28},"中文理解","chinese",{"metricId":40,"metricName":59,"metricCode":60,"maxScore":32,"score":61},"文本写作","writing",8.5,{"metricId":63,"metricName":64,"metricCode":65,"maxScore":32,"score":28},3,"逻辑推理","reasoning",{"metricId":15,"metricName":67,"metricCode":68,"maxScore":32,"score":69},"代码能力","coding",9.5,{"metricId":20,"metricName":71,"metricCode":72,"maxScore":32,"score":73},"多模态","multimodal",7,{"metricId":75,"metricName":76,"metricCode":77,"maxScore":32,"score":28},6,"长文本","long-context",{"metricId":73,"metricName":79,"metricCode":80,"maxScore":32,"score":69},"性价比","cost-performance",[82,84,86,90,92,96,97,100],{"id":20,"name":21,"color":22,"type":18,"toolCount":83},35,{"id":32,"name":33,"color":34,"type":18,"toolCount":85},33,{"id":40,"name":87,"color":88,"type":18,"toolCount":89},"免费增值","#409EFF",21,{"id":15,"name":16,"color":17,"type":18,"toolCount":91},18,{"id":93,"name":94,"color":95,"type":18,"toolCount":32},11,"多模态理解","#9C27B0",{"id":28,"name":29,"color":30,"type":18,"toolCount":32},{"id":63,"name":98,"color":99,"type":18,"toolCount":73},"付费使用","#E6A23C",{"id":24,"name":25,"color":26,"type":18,"toolCount":73}]