[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-image-service-v2-gpt-5.6-luna":3,"text-service-v2-gpt-5.6-luna":205},[4],{"id":5,"aliases":6,"types":7,"pricing_group":10,"pricing":11,"display_name":5,"model_capabilities":28,"model_capabilities_zh":101,"model_data":150,"introduction":193,"introduction_zh":194,"description":195,"description_zh":196,"title_focus":197,"title_focus_zh":198,"keyword_extras":199,"keyword_extras_zh":202},"gpt-5.6-luna",[5],[8,9],"openai","openai\u002Fresponses\u002Fv1","openai\u002Fgpt",{"input_tokens_tiers":12},{"tiers":13},[14,21],{"max_tokens":15,"pricing":16},272000,{"input_tokens":17,"output_tokens":18,"cache_read_tokens":19,"cache_write_tokens":20},"0.2","1.2","0.02","0.25",{"max_tokens":22,"pricing":23},0,{"input_tokens":24,"output_tokens":25,"cache_read_tokens":26,"cache_write_tokens":27},"0.4","1.8","0.04","0.5",{"groups":29,"highlights":91,"overall":95,"scenario":97,"summary":99,"weakness":100},[30,68,82],{"items":31,"name":67},[32,38,43,47,51,55,59,63],{"detail":33,"method":34,"name":35,"samples":36,"score":37},"","Mixed","Agent Collaboration",18,61,{"detail":33,"method":39,"name":40,"samples":41,"score":42},"Objective","Code Generation",12,83,{"detail":33,"method":34,"name":44,"samples":45,"score":46},"Code Comprehension & Debugging",6,80,{"detail":33,"method":48,"name":49,"samples":50,"score":46},"Subjective","Creative Writing",22,{"detail":33,"method":34,"name":52,"samples":53,"score":54},"Information Retrieval \u002F RAG",9,73,{"detail":33,"method":48,"name":56,"samples":57,"score":58},"Long-Context Handling",16,88,{"detail":33,"method":34,"name":60,"samples":61,"score":62},"Math & Science Reasoning",14,79,{"detail":33,"method":48,"name":64,"samples":65,"score":66},"Multilingual Capability",11,86,"Advanced Capabilities",{"items":69,"name":81},[70,74,78],{"detail":33,"method":34,"name":71,"samples":72,"score":73},"General Knowledge",10,93,{"detail":33,"method":48,"name":75,"samples":76,"score":77},"Text Comprehension",23,84,{"detail":33,"method":34,"name":79,"samples":57,"score":80},"Logical Reasoning",72,"General Foundations",{"items":83,"name":90},[84,88],{"detail":33,"method":48,"name":85,"samples":86,"score":87},"Hallucination Control",15,87,{"detail":33,"method":48,"name":89,"samples":36,"score":46},"Instruction Following","Safety & Compliance",[92,93,94],{"name":71,"score":73},{"name":56,"score":58},{"name":85,"score":87},{"rating":96,"score":46},"良好",{"fit":77,"name":98},"Translation & Localization \u002F Scriptwriting \u002F E-commerce Copywriting","GPT-5.6 Luna scored 80 overall, performing well — a fast, cost-efficient member of the GPT-5.6 family. It stands out in general knowledge (93%), long-context handling (88%) and hallucination control (87%), staying accurate over extended documents at a fraction of flagship cost, which suits translation, scriptwriting and e-commerce copy at volume. Comparatively, agent collaboration (61%) is weak; complex tool workflows exceed its reach. High-volume content tasks are where the economics shine.",{"name":35,"note":33,"score":37},{"groups":102,"highlights":141,"overall":145,"scenario":146,"summary":148,"weakness":149},[103,125,134],{"items":104,"name":124},[105,108,111,113,116,118,120,122],{"detail":33,"method":106,"name":107,"samples":36,"score":37},"主\u002F客观混合","智能体协同",{"detail":33,"method":109,"name":110,"samples":41,"score":42},"客观","代码生成",{"detail":33,"method":106,"name":112,"samples":45,"score":46},"代码理解与调试",{"detail":33,"method":114,"name":115,"samples":50,"score":46},"主观","创意写作",{"detail":33,"method":106,"name":117,"samples":53,"score":54},"信息检索 \u002F RAG",{"detail":33,"method":114,"name":119,"samples":57,"score":58},"长文本处理",{"detail":33,"method":106,"name":121,"samples":61,"score":62},"数学与科学推理",{"detail":33,"method":114,"name":123,"samples":65,"score":66},"多语言能力","专项进阶能力",{"items":126,"name":133},[127,129,131],{"detail":33,"method":106,"name":128,"samples":72,"score":73},"通识知识",{"detail":33,"method":114,"name":130,"samples":76,"score":77},"文本理解",{"detail":33,"method":106,"name":132,"samples":57,"score":80},"逻辑推理","通用基础能力",{"items":135,"name":140},[136,138],{"detail":33,"method":114,"name":137,"samples":86,"score":87},"幻觉控制",{"detail":33,"method":114,"name":139,"samples":36,"score":46},"指令遵循","安全与合规能力",[142,143,144],{"name":128,"score":73},{"name":119,"score":58},{"name":137,"score":87},{"rating":96,"score":46},{"fit":77,"name":147},"翻译与本地化 \u002F 短剧剧本生成 \u002F 电商营销文案","GPT-5.6 Luna 综合得分 80，整体表现良好，是GPT-5.6系列中兼顾速度与成本的模型。通识知识（93%）、长文本处理（88%）与幻觉控制（87%）表现突出，在长文档中保持准确且成本远低于旗舰，适合翻译本地化与电商营销文案的大批量生产。相对而言，智能体协同（61%）偏弱，复杂工具工作流超出其射程。大批量内容任务最能体现它的性价比。",{"name":107,"note":33,"score":37},{"bars":151,"metrics":154,"recentAvailability":181,"summary":182},{"degraded":152,"down":153},[],[],[155,158,161,165,169,173,177],{"label":156,"value":157},"Completion Rate","100%",{"label":159,"value":160},"Cache Hit Rate","0.0%",{"label":162,"unit":163,"value":164},"Total Eval Time","s","1811.4",{"label":166,"unit":167,"value":168},"Token Usage","tokens","168,840",{"label":170,"unit":171,"value":172},"Avg. Requests","次","1.0",{"label":174,"unit":175,"value":176},"TPM","tokens\u002Fmin","5,592",{"label":178,"unit":179,"value":180},"QPM","req\u002Fmin","4.9","100.0%",[183,186,189],{"label":184,"unit":163,"value":185},"Latency","7.68",{"label":187,"unit":175,"value":188},"Throughput","5.59K",{"label":190,"unit":191,"value":192},"Availability","%","100.0","GPT-5.6 Luna is the speed-and-cost-efficient model of the GPT-5.6 family — built for batch classification, information extraction, support Q&A and latency-sensitive lightweight workflows.","GPT-5.6 Luna是GPT-5.6系列的速度与成本效率型模型，适合批量分类、信息抽取、客服问答和延迟敏感的轻量工作流。","GPT-5.6 Luna emphasizes fast, cost-efficient responses for high-volume Q&A, information extraction and lightweight automation. Integrate in one click via the SilvaMux unified API.","GPT-5.6 Luna强调快速和高性价比响应，适合高频问答、信息抽取与轻量自动化任务。可通过 SilvaMux 统一 API 一键快速接入。","Fast-Response Model","快速响应模型",[200,201],"high-volume Q&A","information extraction",[203,204],"高频问答","信息抽取",{"ttft":206,"tps":209,"e2e":212,"requests":215,"successfulRequests":216,"successRate":217,"availability":218},{"p50":207,"p95":208},1141,2414,{"p50":210,"p95":211},35,82,{"p50":213,"p95":214},5995,9259,52065,51621,"99.15",[219,219,220,220,219,219,219,219,219,219,219,219,219,219,219,219,219,219,219,219,219,219,219,219],"up","degraded"]