{
  "contentStorage": "postgres",
  "title": "Gemini 3.8 Flash Cyber · Связи",
  "description": "Google DeepMind 于 2026 年 9 月 2 日 发布的 Gemini 3.8 Flash Cyber。",
  "url": "/ru/works/gemini-3-8-flash-cyber/relation.html",
  "entity": {
    "id": "work_gemini_3_8_flash_cyber",
    "type": "work",
    "title": "Gemini 3.8 Flash Cyber",
    "slug": "gemini-3-8-flash-cyber",
    "summary": "Google DeepMind 于 2026 年 9 月 2 日 发布的 Gemini 3.8 Flash Cyber。",
    "status": "published",
    "categories": [
      "category-science-technology"
    ],
    "tags": [],
    "sources": [
      {
        "id": "s1",
        "title": "Introducing Gemini 3.8 Flash and 3.8 Flash Cyber",
        "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/"
      },
      {
        "id": "s2",
        "title": "Gemini 3.8 Flash Cyber | Gemini Enterprise Agent Platform",
        "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-8-flash-cyber"
      },
      {
        "id": "s3",
        "title": "Gemini Enterprise Agent Platform release notes",
        "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/release-notes"
      },
      {
        "id": "s4",
        "title": "OpenAI launches GPT-6 Cyber with stronger exploit research and higher API price than Anthropic and Google",
        "url": "https://venturebeat.com/technology/openai-launches-gpt-6-cyber-with-stronger-exploit-research-and-higher-api-price-than-anthropic-and-google"
      },
      {
        "id": "s5",
        "title": "Google Gemini 3.8 Flash \u0026 3.8 Flash Cyber Benchmarks Explained",
        "url": "https://www.vellum.ai/blog/gemini-3-8-flash-benchmarks-explained"
      }
    ],
    "createdAt": "2026-11-11T13:45:42+08:00",
    "updatedAt": "2026-11-11T13:45:42+08:00",
    "updatedBy": "user:admin",
    "revision": 1,
    "body": "## 发布背景\n\nGemini 3.8 Flash Cyber 是 [[organization_google_deepmind|Google DeepMind]] 于 2026 年 9 月 2 日 发布的 Gemini 模型[^s1]。仅通过 Fairwind Program 向受信任的政府、关键基础设施运营商和软件维护者提供，不面向公开的 Gemini API。[^s1]\n\n## 定价与获取\n\n每百万 token 输入 1.50 美元、输出 7.50 美元（Google Cloud 全球标准价）[^s1]。API 标识为 gemini-3.8-flash-cyber[^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- CyberGym：86.2% Pass@1，厂商自报：高于 GPT-5.5-Cyber 的 85.6% 和前代 Gemini 3.5 Flash Cyber 的 77.5%[^s1]。\n- CWE-Bench：47.2% Pass@1，厂商自报：略低于领先前沿模型的 47.8%，但成本显著更低；该基准由 Collinear 运行[^s1]。\n- Internal benchmark (20 programming languages)：超过 70%，厂商自报：覆盖 20 种编程语言的内部漏洞发现基准，博客称较前代大幅提升[^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与通用版共享同一基础智能，但采用更宽松的网络安全缓解措施，专为漏洞发现与自动修补调优[^s1]。\n\nGoogle Cloud 文档将其列为白名单后的 GA 模型，平台发布日期为 2026 年 9 月 16 日，晚于 9 月 2 日的发布公告[^s1]。\n\n## 定位\n\n它是 [[work_gemini_3_8_flash|Gemini 3.8 Flash]] 的网络安全特化后训练版本，接替 [[work_gemini_3_5_flash_cyber|Gemini 3.5 Flash Cyber]]，属 Gemini 3.x Flash 系列中面向受信任防御者的受限分支，截至核查时未见后继型号公开。截至 2026 年 11 月，该模型处于受限供应状态。",
    "url": "/works/gemini-3-8-flash-cyber.html",
    "lang": "ru",
    "sourceLang": "zh",
    "fallback": true,
    "availableLangs": [
      "zh"
    ],
    "workKind": "ai_model"
  },
  "entities": null,
  "relations": [
    {
      "id": "relation_work_gemini_3_8_flash_cyber_published_by_organization_google_deepmind",
      "fromId": "work_gemini_3_8_flash_cyber",
      "relationType": "published_by",
      "toId": "organization_google_deepmind",
      "sourceRefs": [
        {
          "entityId": "work_gemini_3_8_flash_cyber",
          "sourceId": "s1"
        }
      ],
      "label": "Published by",
      "targetId": "organization_google_deepmind",
      "targetTitle": "Google DeepMind",
      "targetUrl": "/ru/organizations/google-deepmind.html",
      "incoming": false
    }
  ],
  "relationGroups": [
    {
      "type": "published_by",
      "label": "Published by",
      "items": [
        {
          "id": "relation_work_gemini_3_8_flash_cyber_published_by_organization_google_deepmind",
          "fromId": "work_gemini_3_8_flash_cyber",
          "relationType": "published_by",
          "toId": "organization_google_deepmind",
          "sourceRefs": [
            {
              "entityId": "work_gemini_3_8_flash_cyber",
              "sourceId": "s1"
            }
          ],
          "label": "Published by",
          "targetId": "organization_google_deepmind",
          "targetTitle": "Google DeepMind",
          "targetUrl": "/ru/organizations/google-deepmind.html",
          "incoming": false
        }
      ]
    }
  ],
  "lang": "ru",
  "availableLangs": [
    "zh"
  ],
  "fallback": true,
  "infoFields": [
    {
      "key": "work_kind",
      "label": "Work kind",
      "value": "Модель ИИ"
    }
  ],
  "pagination": {
    "page": 1,
    "pageSize": 50,
    "total": 1,
    "pageCount": 1
  },
  "relationsTotal": 1
}