{
  "contentStorage": "postgres",
  "title": "Gemini 3.5 Flash",
  "description": "Google DeepMind 于 2026 年 5 月 19 日 发布的 Gemini 3.5 Flash。",
  "url": "/ko/works/gemini-3-5-flash.html",
  "entity": {
    "id": "work_gemini_3_5_flash",
    "type": "work",
    "title": "Gemini 3.5 Flash",
    "slug": "gemini-3-5-flash",
    "summary": "Google DeepMind 于 2026 年 5 月 19 日 发布的 Gemini 3.5 Flash。",
    "status": "published",
    "categories": [
      "category-science-technology"
    ],
    "tags": [],
    "sources": [
      {
        "id": "s1",
        "title": "Gemini 3.5 Flash - Model Card — Google DeepMind",
        "url": "https://deepmind.google/models/model-cards/gemini-3-5-flash/"
      },
      {
        "id": "s2",
        "title": "Google I/O 2026: News and announcements",
        "url": "https://blog.google/innovation-and-ai/technology/developers-tools/google-io-2026-collection/"
      },
      {
        "id": "s3",
        "title": "Gemini (language model) — Wikipedia",
        "url": "https://en.wikipedia.org/wiki/Gemini_(language_model)"
      },
      {
        "id": "s4",
        "title": "Gemini 3.6 Flash - Model Card — Google DeepMind",
        "url": "https://deepmind.google/models/model-cards/gemini-3-6-flash/"
      },
      {
        "id": "s5",
        "title": "Models — Gemini API",
        "url": "https://ai.google.dev/gemini-api/docs/models"
      }
    ],
    "createdAt": "2026-11-11T13:45:13+08:00",
    "updatedAt": "2026-11-11T13:45:13+08:00",
    "updatedBy": "user:admin",
    "revision": 1,
    "body": "## 发布背景\n\nGemini 3.5 Flash 是 [[organization_google_deepmind|Google DeepMind]] 于 2026 年 5 月 19 日 发布的 Gemini 模型[^s1]。2026年5月19日在Google I/O 2026发布，是Gemini 3.5家族的首款模型，发布时成为搜索AI Mode的全球默认模型。[^s1]\n\n## 定价与获取\n\n每百万 token 输入 1.50 美元、输出 9.00 美元[^s1]。API 标识为 gemini-3.5-flash[^s1]。\n\n## 性能表现\n\n以下均为 Google DeepMind 在发布公告与技术报告中公布的厂商自报成绩，尚未经独立复现：\n\n- Terminal-Bench 2.1：76.2%，厂商自报；高于Gemini 3.1 Pro的70.3%，低于GPT-5.5的78.2%[^s1]。\n- MCP Atlas：83.6%，厂商自报；领先同场模型，高于Gemini 3.1 Pro的78.2%[^s1]。\n- GDPval-AA：1656 Elo，厂商自报；低于GPT-5.5的1769 Elo和Claude Opus 4.7的1753 Elo[^s1]。\n- CharXiv Reasoning：84.2%，厂商自报；略高于GPT-5.5的84.1%[^s1]。\n\n## 主要特点\n\n上下文窗口 1,048,576 token，最大输出 65,536 token[^s1]。\n\n该模型基于[[work_gemini_3_flash|Gemini 3 Flash]]推理基础，谷歌称其输出速度比其他前沿模型快约4倍，主打智能体与编码工作流[^s1]。\n\n截至2026年11月10日，Gemini API模型页已不再列出该模型，API请求转至Gemini 3.6 Flash，Vertex AI标注其退役不早于2027年5月19日[^s1]。\n\n## 定位\n\n它是Gemini 3 Flash的迭代版本，发布时接替Gemini 3.1 Pro成为旗舰级Flash模型，后由2026年7月21日发布的Gemini 3.6 Flash接替。截至 2026 年 11 月，该模型处于受限供应状态。",
    "url": "/works/gemini-3-5-flash.html",
    "lang": "ko",
    "sourceLang": "zh",
    "fallback": true,
    "availableLangs": [
      "zh"
    ],
    "workKind": "ai_model"
  },
  "entities": null,
  "relations": [
    {
      "id": "relation_work_gemini_3_5_flash_published_by_organization_google_deepmind",
      "fromId": "work_gemini_3_5_flash",
      "relationType": "published_by",
      "toId": "organization_google_deepmind",
      "sourceRefs": [
        {
          "entityId": "work_gemini_3_5_flash",
          "sourceId": "s1"
        }
      ],
      "label": "Published by",
      "targetId": "organization_google_deepmind",
      "targetTitle": "Google DeepMind",
      "targetUrl": "/ko/organizations/google-deepmind.html",
      "incoming": false
    }
  ],
  "mentions": [
    {
      "fromId": "work_gemini_3_5_flash_cyber",
      "fromTitle": "Gemini 3.5 Flash Cyber",
      "fromUrl": "/ko/works/gemini-3-5-flash-cyber.html",
      "fromType": "work"
    },
    {
      "fromId": "work_gemini_3_5_flash_lite",
      "fromTitle": "Gemini 3.5 Flash-Lite",
      "fromUrl": "/ko/works/gemini-3-5-flash-lite.html",
      "fromType": "work"
    },
    {
      "fromId": "work_gemini_3_6_flash",
      "fromTitle": "Gemini 3.6 Flash",
      "fromUrl": "/ko/works/gemini-3-6-flash.html",
      "fromType": "work"
    }
  ],
  "versions": [
    {
      "revision": 1,
      "updatedAt": "2026-11-11T13:45:13+08:00",
      "message": "当前版本",
      "hasSnapshot": true,
      "url": "/ko/works/gemini-3-5-flash/at/1.html"
    }
  ],
  "lang": "ko",
  "availableLangs": [
    "zh"
  ],
  "fallback": true,
  "infoFields": [
    {
      "key": "work_kind",
      "label": "Work kind",
      "value": "인공지능 모델"
    }
  ],
  "relationsTotal": 1,
  "relationCounts": [
    {
      "relationType": "published_by",
      "incoming": false,
      "label": "Published by",
      "total": 1
    }
  ],
  "mentionsTotal": 3
}