{
  "contentStorage": "postgres",
  "title": "Gemini 3 Pro · ความสัมพันธ์",
  "description": "Google DeepMind 于 2025 年 11 月 18 日 发布的 Gemini 3 Pro。",
  "url": "/th/works/gemini-3-pro/relation.html",
  "entity": {
    "id": "work_gemini_3_pro",
    "type": "work",
    "title": "Gemini 3 Pro",
    "slug": "gemini-3-pro",
    "summary": "Google DeepMind 于 2025 年 11 月 18 日 发布的 Gemini 3 Pro。",
    "status": "published",
    "categories": [
      "category-science-technology"
    ],
    "tags": [],
    "sources": [
      {
        "id": "s1",
        "title": "A new era of intelligence with Gemini 3",
        "url": "https://blog.google/products/gemini/gemini-3/"
      },
      {
        "id": "s2",
        "title": "Gemini (language model) - Wikipedia",
        "url": "https://en.wikipedia.org/wiki/Gemini_(language_model)"
      },
      {
        "id": "s3",
        "title": "Gemini API Models - Google AI for Developers",
        "url": "https://ai.google.dev/gemini-api/docs/models"
      },
      {
        "id": "s4",
        "title": "Gemini Developer API pricing - Google AI for Developers",
        "url": "https://ai.google.dev/gemini-api/docs/pricing"
      },
      {
        "id": "s5",
        "title": "Gemini 3 Deep Think is now available",
        "url": "https://blog.google/products-and-platforms/products/gemini/gemini-3-deep-think/"
      }
    ],
    "createdAt": "2026-11-11T13:44:51+08:00",
    "updatedAt": "2026-11-11T13:44:51+08:00",
    "updatedBy": "user:admin",
    "revision": 1,
    "body": "## 发布背景\n\nGemini 3 Pro 是 [[organization_google_deepmind|Google DeepMind]] 于 2025 年 11 月 18 日 发布的 Gemini 模型[^s1]。2025年11月18日以预览版发布，是 Gemini 3 系列的基础模型，采用稀疏混合专家架构、原生多模态推理。[^s1]\n\n## 定价与获取\n\n每百万 token 输入 2 美元、输出 12 美元（提示不超过 20 万 token；超过则输入 4 美元、输出 18 美元）[^s4]。API 标识为 gemini-3-pro-preview[^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- LMArena：1501 Elo，厂商自报，居榜首[^s1]。\n- GPQA Diamond：91.9%，厂商自报，高于 Gemini 2.5 Pro[^s1]。\n- Humanity's Last Exam（无工具）：37.5%，厂商自报，高于 Gemini 2.5 Pro[^s1]。\n- SWE-bench Verified：76.2%，厂商自报，显著高于 Gemini 2.5 Pro[^s1]。\n\n## 主要特点\n\n上下文窗口 1,000,000 token，最大输出 64,000 token[^s1]。\n\n发布当日即进入 Gemini 应用、搜索 AI Mode、AI Studio 与 Vertex AI，始终未转为正式版[^s1]。\n\n原预览端点于 2026 年 3 月 9 日停用，由 2026 年 2 月 19 日发布的 Gemini 3.1 Pro 接替[^s1]。\n\nDeep Think 是构建于 Gemini 3 Pro 之上的增强推理模式，并非独立模型[^s5]：2025 年 12 月 4 日起向 Google AI Ultra 订阅者开放，使用时需在 Gemini 应用中同时选择 Deep Think 开关与 Gemini 3 Pro 模型；官方 Gemini API 模型目录中没有 Deep Think 条目或公开模型 ID[^s3]。\n\n## 定位\n\n承接 [[work_gemini_2_5_pro|Gemini 2.5 Pro]] 的旗舰位置，为 Gemini 3 系列首款模型，后由 Gemini 3.1 Pro 接替。该模型已于 2026 年 3 月 9 日退役。",
    "url": "/works/gemini-3-pro.html",
    "lang": "th",
    "sourceLang": "zh",
    "fallback": true,
    "availableLangs": [
      "zh"
    ],
    "workKind": "ai_model"
  },
  "entities": null,
  "relations": [
    {
      "id": "relation_work_gemini_3_pro_published_by_organization_google_deepmind",
      "fromId": "work_gemini_3_pro",
      "relationType": "published_by",
      "toId": "organization_google_deepmind",
      "sourceRefs": [
        {
          "entityId": "work_gemini_3_pro",
          "sourceId": "s1"
        }
      ],
      "label": "Published by",
      "targetId": "organization_google_deepmind",
      "targetTitle": "Google DeepMind",
      "targetUrl": "/th/organizations/google-deepmind.html",
      "incoming": false
    }
  ],
  "mentions": [
    {
      "fromId": "work_gemini_3_1_flash_lite",
      "fromTitle": "Gemini 3.1 Flash-Lite",
      "fromUrl": "/th/works/gemini-3-1-flash-lite.html",
      "fromType": "work"
    },
    {
      "fromId": "work_gemini_3_1_pro",
      "fromTitle": "Gemini 3.1 Pro",
      "fromUrl": "/th/works/gemini-3-1-pro.html",
      "fromType": "work"
    },
    {
      "fromId": "work_gemini_3_flash",
      "fromTitle": "Gemini 3 Flash",
      "fromUrl": "/th/works/gemini-3-flash.html",
      "fromType": "work"
    }
  ],
  "relationGroups": [
    {
      "type": "published_by",
      "label": "Published by",
      "items": [
        {
          "id": "relation_work_gemini_3_pro_published_by_organization_google_deepmind",
          "fromId": "work_gemini_3_pro",
          "relationType": "published_by",
          "toId": "organization_google_deepmind",
          "sourceRefs": [
            {
              "entityId": "work_gemini_3_pro",
              "sourceId": "s1"
            }
          ],
          "label": "Published by",
          "targetId": "organization_google_deepmind",
          "targetTitle": "Google DeepMind",
          "targetUrl": "/th/organizations/google-deepmind.html",
          "incoming": false
        }
      ]
    }
  ],
  "lang": "th",
  "availableLangs": [
    "zh"
  ],
  "fallback": true,
  "infoFields": [
    {
      "key": "work_kind",
      "label": "Work kind",
      "value": "โมเดล AI"
    }
  ],
  "pagination": {
    "page": 1,
    "pageSize": 50,
    "total": 1,
    "pageCount": 1
  },
  "relationsTotal": 1,
  "mentionsTotal": 3
}