[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-image-service-v2-gemini-3.5-flash":3,"text-service-v2-gemini-3.5-flash":199},[4],{"id":5,"aliases":6,"types":7,"pricing_group":9,"pricing":10,"display_name":5,"capabilities":15,"model_capabilities":20,"model_capabilities_zh":93,"model_data":142,"introduction":187,"introduction_zh":188,"description":189,"description_zh":190,"title_focus":191,"title_focus_zh":192,"keyword_extras":193,"keyword_extras_zh":196},"gemini-3.5-flash",[5],[8],"openai","google\u002Fgemini",{"flat_tokens":11},{"input_tokens":12,"output_tokens":13,"cache_read_tokens":14,"cache_write_tokens":12},"0.3","2.5","0.075",{"context_window":16,"max_output":17,"modalities":18},1048576,65536,[19],"text",{"groups":21,"highlights":82,"overall":86,"scenario":88,"summary":91,"weakness":92},[22,59,72],{"items":23,"name":58},[24,30,35,39,43,46,50,54],{"detail":25,"method":26,"name":27,"samples":28,"score":29},"","Mixed","Agent Collaboration",18,67,{"detail":25,"method":31,"name":32,"samples":33,"score":34},"Objective","Code Generation",12,42,{"detail":25,"method":26,"name":36,"samples":37,"score":38},"Code Comprehension & Debugging",6,64,{"detail":25,"method":26,"name":40,"samples":41,"score":42},"Creative Writing",22,60,{"detail":25,"method":26,"name":44,"samples":45,"score":29},"Information Retrieval \u002F RAG",9,{"detail":25,"method":26,"name":47,"samples":48,"score":49},"Long-Context Handling",16,78,{"detail":25,"method":31,"name":51,"samples":52,"score":53},"Math & Science Reasoning",14,77,{"detail":25,"method":26,"name":55,"samples":56,"score":57},"Multilingual Capability",11,82,"Advanced Capabilities",{"items":60,"name":71},[61,65,69],{"detail":25,"method":26,"name":62,"samples":63,"score":64},"General Knowledge",10,92,{"detail":25,"method":26,"name":66,"samples":67,"score":68},"Text Comprehension",23,61,{"detail":25,"method":26,"name":70,"samples":48,"score":29},"Logical Reasoning","General Foundations",{"items":73,"name":81},[74,78],{"detail":25,"method":26,"name":75,"samples":76,"score":77},"Hallucination Control",15,70,{"detail":25,"method":79,"name":80,"samples":28,"score":57},"Subjective","Instruction Following","Safety & Compliance",[83,84,85],{"name":62,"score":64},{"name":55,"score":57},{"name":80,"score":57},{"rating":87,"score":77},"良好",{"fit":89,"name":90},74,"Customer Support \u002F Enterprise Knowledge Q&A \u002F Translation & Localization","Gemini 3.5 Flash scored 70 overall, performing well — an efficient all-rounder for scale. It stands out in general knowledge (92%), multilingual capability (82%) and instruction following (82%), answering reliably across languages while obeying format constraints, which fits customer support and knowledge-base Q&A at high request volume. Comparatively, code generation (42%) is weak; coding tasks should go elsewhere. For fast, well-behaved knowledge responses it is a sensible default.",{"name":32,"note":25,"score":34},{"groups":94,"highlights":133,"overall":137,"scenario":138,"summary":140,"weakness":141},[95,116,125],{"items":96,"name":115},[97,100,103,105,107,109,111,113],{"detail":25,"method":98,"name":99,"samples":28,"score":29},"主\u002F客观混合","智能体协同",{"detail":25,"method":101,"name":102,"samples":33,"score":34},"客观","代码生成",{"detail":25,"method":98,"name":104,"samples":37,"score":38},"代码理解与调试",{"detail":25,"method":98,"name":106,"samples":41,"score":42},"创意写作",{"detail":25,"method":98,"name":108,"samples":45,"score":29},"信息检索 \u002F RAG",{"detail":25,"method":98,"name":110,"samples":48,"score":49},"长文本处理",{"detail":25,"method":101,"name":112,"samples":52,"score":53},"数学与科学推理",{"detail":25,"method":98,"name":114,"samples":56,"score":57},"多语言能力","专项进阶能力",{"items":117,"name":124},[118,120,122],{"detail":25,"method":98,"name":119,"samples":63,"score":64},"通识知识",{"detail":25,"method":98,"name":121,"samples":67,"score":68},"文本理解",{"detail":25,"method":98,"name":123,"samples":48,"score":29},"逻辑推理","通用基础能力",{"items":126,"name":132},[127,129],{"detail":25,"method":98,"name":128,"samples":76,"score":77},"幻觉控制",{"detail":25,"method":130,"name":131,"samples":28,"score":57},"主观","指令遵循","安全与合规能力",[134,135,136],{"name":119,"score":64},{"name":114,"score":57},{"name":131,"score":57},{"rating":87,"score":77},{"fit":89,"name":139},"智能客服 \u002F 企业知识库问答 \u002F 翻译与本地化","Gemini 3.5 Flash 综合得分 70，整体表现良好，是一款面向规模化应用的高效模型。通识知识（92%）、多语言能力（82%）与指令遵循（82%）表现突出，跨语言回答可靠且严格遵守格式约束，适合高请求量的智能客服与知识库问答。相对而言，代码生成（42%）偏弱，编码任务建议交给其他模型。快速守规的知识问答场景选它是合理的默认。",{"name":102,"note":25,"score":34},{"bars":143,"metrics":148,"recentAvailability":175,"summary":176},{"degraded":144,"down":146},[145],1,[147],0,[149,152,155,159,163,167,171],{"label":150,"value":151},"Completion Rate","99%",{"label":153,"value":154},"Cache Hit Rate","0.0%",{"label":156,"unit":157,"value":158},"Total Eval Time","s","9630.5",{"label":160,"unit":161,"value":162},"Token Usage","tokens","240,662",{"label":164,"unit":165,"value":166},"Avg. Requests","次","1.0",{"label":168,"unit":169,"value":170},"TPM","tokens\u002Fmin","1,499",{"label":172,"unit":173,"value":174},"QPM","req\u002Fmin","0.9","99.33%",[177,180,183],{"label":178,"unit":157,"value":179},"Latency","13.85",{"label":181,"unit":169,"value":182},"Throughput","1.5K",{"label":184,"unit":185,"value":186},"Availability","%","99.33","Gemini 3.5 Flash is a multimodal model built for scale, strengthening coding and agentic capability on top of fast responses — suited to real-time assistants, code workflows and high-concurrency content processing.","Gemini 3.5 Flash是一款面向规模化应用的多模态模型，在快速响应之外强化了编程和智能体能力，适合实时助手、代码工作流和高并发内容处理。","Gemini 3.5 Flash balances speed, coding and agentic execution for high-frequency multimodal processing and continuous tool calls. Integrate in one click via the SilvaMux unified API.","Gemini 3.5 Flash兼顾速度、编程与智能体执行，适合高频多模态处理和持续工具调用任务。可通过 SilvaMux 统一 API 一键快速接入。","Multimodal Agent Model","多模态智能体模型",[194,195],"fast response","coding tasks",[197,198],"快速响应","代码任务",{"ttft":200,"tps":203,"e2e":205,"requests":208,"successfulRequests":209,"successRate":210,"availability":211},{"p50":201,"p95":202},2151,2478,{"p50":34,"p95":204},85,{"p50":206,"p95":207},6887,7709,163992,159321,"97.15",[212,212,212,212,212,212,212,213,212,212,212,212,212,212,212,212,212,212,212,212,212,212,212,212],"up","degraded"]