[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-image-service-v2-gpt-5.4":3,"text-service-v2-gpt-5.4":202},[4],{"id":5,"aliases":6,"types":7,"pricing_group":10,"pricing":11,"display_name":5,"model_capabilities":26,"model_capabilities_zh":100,"introduction":149,"introduction_zh":150,"description":151,"description_zh":152,"title_focus":153,"title_focus_zh":154,"keyword_extras":155,"keyword_extras_zh":158,"model_data":161},"gpt-5.4",[5],[8,9],"openai","openai\u002Fresponses\u002Fv1","openai\u002Fgpt",{"input_tokens_tiers":12},{"tiers":13},[14,20],{"max_tokens":15,"pricing":16},270000,{"input_tokens":17,"output_tokens":18,"cache_read_tokens":19,"cache_write_tokens":17},"2.5","15","0.25",{"max_tokens":21,"pricing":22},0,{"input_tokens":23,"output_tokens":24,"cache_read_tokens":25,"cache_write_tokens":23},"5","22.5","0.5",{"groups":27,"highlights":89,"overall":93,"scenario":95,"summary":98,"weakness":99},[28,66,79],{"items":29,"name":65},[30,36,41,45,49,53,57,61],{"detail":31,"method":32,"name":33,"samples":34,"score":35},"","Mixed","Agent Collaboration",18,54,{"detail":31,"method":37,"name":38,"samples":39,"score":40},"Objective","Code Generation",12,79,{"detail":31,"method":32,"name":42,"samples":43,"score":44},"Code Comprehension & Debugging",6,77,{"detail":31,"method":46,"name":47,"samples":48,"score":40},"Subjective","Creative Writing",22,{"detail":31,"method":32,"name":50,"samples":51,"score":52},"Information Retrieval \u002F RAG",9,82,{"detail":31,"method":32,"name":54,"samples":55,"score":56},"Long-Context Handling",16,81,{"detail":31,"method":32,"name":58,"samples":59,"score":60},"Math & Science Reasoning",14,55,{"detail":31,"method":46,"name":62,"samples":63,"score":64},"Multilingual Capability",11,87,"Advanced Capabilities",{"items":67,"name":78},[68,72,75],{"detail":31,"method":32,"name":69,"samples":70,"score":71},"General Knowledge",10,96,{"detail":31,"method":46,"name":73,"samples":74,"score":44},"Text Comprehension",23,{"detail":31,"method":32,"name":76,"samples":55,"score":77},"Logical Reasoning",61,"General Foundations",{"items":80,"name":88},[81,85],{"detail":31,"method":46,"name":82,"samples":83,"score":84},"Hallucination Control",15,88,{"detail":31,"method":46,"name":86,"samples":34,"score":87},"Instruction Following",86,"Safety & Compliance",[90,91,92],{"name":69,"score":71},{"name":82,"score":84},{"name":62,"score":64},{"rating":94,"score":44},"良好",{"fit":96,"name":97},84,"Translation & Localization \u002F E-commerce Copywriting \u002F Scriptwriting","GPT-5.4 scored 77 overall, performing well — a knowledge-first generalist. It stands out in general knowledge (96%), hallucination control (88%) and multilingual capability (87%), combining broad accurate coverage with careful, low-hallucination output, which serves localization and marketing content well. Comparatively, agent collaboration (54%) is weak; tool-orchestration workflows are a stretch. Knowledge-accuracy-first content tasks are where it fits.",{"name":33,"note":31,"score":35},{"groups":101,"highlights":140,"overall":144,"scenario":145,"summary":147,"weakness":148},[102,124,133],{"items":103,"name":123},[104,107,110,112,115,117,119,121],{"detail":31,"method":105,"name":106,"samples":34,"score":35},"主\u002F客观混合","智能体协同",{"detail":31,"method":108,"name":109,"samples":39,"score":40},"客观","代码生成",{"detail":31,"method":105,"name":111,"samples":43,"score":44},"代码理解与调试",{"detail":31,"method":113,"name":114,"samples":48,"score":40},"主观","创意写作",{"detail":31,"method":105,"name":116,"samples":51,"score":52},"信息检索 \u002F RAG",{"detail":31,"method":105,"name":118,"samples":55,"score":56},"长文本处理",{"detail":31,"method":105,"name":120,"samples":59,"score":60},"数学与科学推理",{"detail":31,"method":113,"name":122,"samples":63,"score":64},"多语言能力","专项进阶能力",{"items":125,"name":132},[126,128,130],{"detail":31,"method":105,"name":127,"samples":70,"score":71},"通识知识",{"detail":31,"method":113,"name":129,"samples":74,"score":44},"文本理解",{"detail":31,"method":105,"name":131,"samples":55,"score":77},"逻辑推理","通用基础能力",{"items":134,"name":139},[135,137],{"detail":31,"method":113,"name":136,"samples":83,"score":84},"幻觉控制",{"detail":31,"method":113,"name":138,"samples":34,"score":87},"指令遵循","安全与合规能力",[141,142,143],{"name":127,"score":71},{"name":136,"score":84},{"name":122,"score":64},{"rating":94,"score":44},{"fit":96,"name":146},"翻译与本地化 \u002F 电商营销文案 \u002F 短剧剧本生成","GPT-5.4 综合得分 77，整体表现良好，是一款知识优先的通用模型。通识知识（96%）、幻觉控制（88%）与多语言能力（87%）表现突出，知识广博且输出谨慎、低幻觉，在翻译本地化与营销内容场景中优势明显。相对而言，智能体协同（54%）偏弱，工具编排类工作流有些勉强。知识准确性优先的内容任务最适合它。",{"name":106,"note":31,"score":35},"GPT-5.4 is OpenAI's general-purpose model with broad accurate knowledge, steady factual grounding and multilingual strength — suited to localization, e-commerce copy and scriptwriting.","GPT-5.4是OpenAI的通用模型，通识知识覆盖广、事实准确性和多语言能力稳定，适合翻译本地化、电商营销文案和短剧剧本生成等内容型任务。","GPT-5.4 strengthens knowledge accuracy and multilingual generation for content creation, translation and everyday knowledge Q&A. Integrate in one click via the SilvaMux unified API.","GPT-5.4强化知识准确性与多语言生成，适合内容创作、翻译与日常知识问答。可通过 SilvaMux 统一 API 一键快速接入。","General Knowledge Model","通用知识模型",[156,157],"multilingual","content creation",[159,160],"多语言","内容创作",{"summary":162,"recentAvailability":175,"observedAt":31,"bars":176,"metrics":180},[163,167,171],{"label":164,"value":165,"unit":166},"modelsDetail.performance.latency","12.85","s",{"label":168,"value":169,"unit":170},"modelsDetail.performance.throughput","54.40K","tokens\u002Fmin",{"label":172,"value":173,"unit":174},"modelsDetail.performance.availability","98.9","%","98.9%",{"degraded":177,"down":179},[178],3,[],[181,184,187,191,195,198],{"label":182,"value":183},"modelsDetail.performance.completionRate","96%",{"label":185,"value":186},"modelsDetail.performance.cacheHitRate","83%",{"label":188,"value":189,"unit":190},"modelsDetail.performance.tokenUsage","2,787,772,078","tokens",{"label":192,"value":193,"unit":194},"modelsDetail.performance.avgRequests","39","modelsDetail.performance.requestsUnit",{"label":196,"value":197,"unit":170},"modelsDetail.performance.tpm","54,401",{"label":199,"value":200,"unit":201},"modelsDetail.performance.qpm","989","req\u002Fmin",{"ttft":203,"tps":206,"e2e":209,"requests":212,"successfulRequests":213,"successRate":214,"availability":215},{"p50":204,"p95":205},1858,2130,{"p50":207,"p95":208},98,100,{"p50":210,"p95":211},8830,8836,128469,126894,"98.77",[216,216,216,216,216,216,216,216,216,217,216,216,216,216,216,216,216,216,216,216,216,216,216,217],"up","degraded"]