[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-doubao-seed-2.0-lite":3},[4],{"id":5,"aliases":6,"types":7,"pricing":9,"capabilities":30,"model_capabilities":38,"model_data":110,"introduction":153,"description":154,"title_focus":155,"keyword_extras":156},"doubao-seed-2.0-lite",[5],[8],"volcengine\u002Fchat\u002Fv3",{"input_tokens_tiers":10},{"tiers":11},[12,18,24],{"max_tokens":13,"pricing":14},32000,{"input_tokens":15,"output_tokens":16,"cache_read_tokens":17,"cache_write_tokens":15},"0.6","3.6","0.12",{"max_tokens":19,"pricing":20},128000,{"input_tokens":21,"output_tokens":22,"cache_read_tokens":23,"cache_write_tokens":21},"0.9","5.4","0.18",{"max_tokens":25,"pricing":26},0,{"input_tokens":27,"output_tokens":28,"cache_read_tokens":29,"cache_write_tokens":27},"1.8","10.8","0.36",{"context_window":31,"released_at":32,"modalities":33},256000,"2026-02-14",[34,35,36,37],"text","image","video","audio",{"groups":39,"highlights":99,"overall":103,"scenario":106,"summary":108,"weakness":109},[40,77,90],{"items":41,"name":76},[42,48,53,58,61,65,69,73],{"detail":43,"method":44,"name":45,"samples":46,"score":47},"","主\u002F客观混合","智能体协同",7,40,{"detail":43,"method":49,"name":50,"samples":51,"score":52},"客观","代码生成",10,80,{"detail":43,"method":54,"name":55,"samples":56,"score":57},"主观","代码理解与调试",12,74,{"detail":43,"method":54,"name":59,"samples":60,"score":52},"创意写作",17,{"detail":43,"method":44,"name":62,"samples":63,"score":64},"信息检索 \u002F RAG",13,94,{"detail":43,"method":54,"name":66,"samples":67,"score":68},"长文本处理",3,87,{"detail":43,"method":49,"name":70,"samples":71,"score":72},"数学与科学推理",14,86,{"detail":43,"method":44,"name":74,"samples":51,"score":75},"多语言能力",84,"专项进阶能力",{"items":78,"name":89},[79,82,86],{"detail":43,"method":44,"name":80,"samples":81,"score":57},"通识知识",15,{"detail":43,"method":54,"name":83,"samples":84,"score":85},"文本理解",5,81,{"detail":43,"method":44,"name":87,"samples":88,"score":52},"逻辑推理",20,"通用基础能力",{"items":91,"name":98},[92,96],{"detail":43,"method":54,"name":93,"samples":94,"score":95},"幻觉控制",9,77,{"detail":43,"method":54,"name":97,"samples":71,"score":75},"指令遵循","安全与合规能力",[100,101,102],{"name":62,"score":64},{"name":66,"score":68},{"name":70,"score":72},{"rating":104,"score":105},"良好",79,{"fit":105,"name":107},"代码研发 \u002F 信息检索 \u002F 企业知识库","Doubao Seed 2.0 Lite在本轮能力评测中综合得分79，整体表现良好。其信息检索 \u002F RAG（94%）和长文本处理（87%）表现突出，能够从知识库和多来源资料中定位信息、组织证据并生成有依据的回答，并连续理解长篇材料并完成跨段落信息整合。对企业知识库团队、AI 应用开发者和研究人员而言，它适合用于资料检索、文档问答、信息抽取和检索增强生成场景，解决从大量资料中找到相关信息，并形成可用于后续决策的回答；数学与科学推理（86%）也为相关任务提供补充。在能力边界上，智能体协同（40%）相对较弱，在需要多步骤自主规划、连续工具调用或跨智能体协作的任务中，建议增加人工复核或专项验证。",{"name":45,"note":43,"score":47},{"bars":111,"metrics":114,"recentAvailability":141,"summary":142},{"degraded":112,"down":113},[],[],[115,118,121,125,129,133,137],{"label":116,"value":117},"完成率","100%",{"label":119,"value":120},"缓存命中率","动态变化",{"label":122,"unit":123,"value":124},"累计评测耗时","s","6890.3",{"label":126,"unit":127,"value":128},"Token 消耗","tokens","327,254",{"label":130,"unit":131,"value":132},"平均请求次数","次","1.0",{"label":134,"unit":135,"value":136},"TPM","tokens\u002Fmin","2,850",{"label":138,"unit":139,"value":140},"QPM","req\u002Fmin","1.3","100.0%",[143,146,149],{"label":144,"unit":123,"value":145},"延迟","36.55",{"label":147,"unit":135,"value":148},"吞吐量","2.85K",{"label":150,"unit":151,"value":152},"可用率","%","100.0","Doubao Seed 2.0 Lite是一款侧重信息检索\u002FRAG的模型，在长文本处理方面也有较好表现。它适合处理企业知识检索、资料问答与检索增强生成，可作为相关业务流程中的理解、推理或生成能力底座。","Doubao Seed 2.0 Lite具备信息检索\u002FRAG、长文本处理、数学与科学推理能力，可完成知识库检索、资料问答和证据整合，并支持长篇材料理解和跨段落信息整合。面向企业知识库团队、AI 应用开发者和研究人员，适合资料检索、文档问答、信息抽取和检索增强生成；可通过 SilvaMux 统一 API 一键快速接入。","长文检索模型",[157,66],"信息检索与RAG"]