{
  "contentStorage": "postgres",
  "title": "GPT-6.1 Terra · Relasi",
  "description": "GPT-6.1 Terra 是 OpenAI 于 2026 年 10 月 26 日发布的 GPT-6 世代中端高效模型，主打视觉推理与电脑操作。",
  "url": "/ms/works/gpt-6-1-terra/relation.html",
  "entity": {
    "id": "work_gpt_6_1_terra",
    "type": "work",
    "title": "GPT-6.1 Terra",
    "slug": "gpt-6-1-terra",
    "aliases": [
      "gpt-6.1-terra"
    ],
    "summary": "GPT-6.1 Terra 是 OpenAI 于 2026 年 10 月 26 日发布的 GPT-6 世代中端高效模型，主打视觉推理与电脑操作。",
    "status": "published",
    "categories": [
      "category-science-technology"
    ],
    "tags": [],
    "sources": [
      {
        "id": "s1",
        "title": "Introducing GPT-6.1 Terra | OpenAI",
        "url": "https://openai.com/index/introducing-gpt-6-1-terra/"
      },
      {
        "id": "s2",
        "title": "GPT-6.1 Terra Model | OpenAI API",
        "url": "https://developers.openai.com/api/docs/models/gpt-6.1-terra"
      },
      {
        "id": "s3",
        "title": "Using GPT-6 | OpenAI API",
        "url": "https://developers.openai.com/api/docs/guides/latest-model"
      }
    ],
    "createdAt": "2026-11-10T15:07:10+08:00",
    "updatedAt": "2026-11-10T15:07:10+08:00",
    "updatedBy": "user:admin",
    "revision": 1,
    "body": "GPT-6.1 Terra 是 OpenAI 于 2026 年 10 月 26 日发布的 GPT-6 世代中端高效模型，主打视觉推理与电脑操作。[^s1]\n\n## 发布背景\n发布时上下文窗口为1,050,000 tokens，最大输出为128,000 tokens，API 标识为 gpt-6.1-terra。[^s1]\n主打详细视觉推理、精确电脑操作与清晰自然写作，知识截止 2026-06-01。[^s1]\n\n## 定价与获取\n每百万 token 输入 1 美元、输出 5 美元。[^s1]\n\n## 性能表现\n以下为OpenAI公布的自家评测结果，尚未经独立复现。[^s1]\n- Agents' Last Exam：比 GPT-5.6 Terra 高 15 个百分点（中等推理 effort）（与 GPT-6 Astra 持平，仅为其十分之一 token 单价，成本约为前代三分之一）[^s1]\n- Terminal-Bench 4.0：50.3%（最高推理 effort）（GPT-5.6 Terra 为 21.5%，翻倍以上；最复杂编码仍推荐 GPT-6.1 Sol 或 GPT-6 Astra）[^s1]\n- MMMU-Pro：在所有推理设置下均优于 GPT-5.6 Terra，接近 GPT-6 Astra 与 GPT-6.1 Sol（与 GPT-5.6 Terra、GPT-6 Astra 及 GPT-6.1 Sol 对照）[^s1]\n- 内部易错提示事实性测试：低推理 effort 错误率 16.7%，最高 effort 降至 4.3%（GPT-5.6 Terra 低 effort 错误率为 53.6%，最高 effort 下降 18 个百分点）[^s1]\n\n## 主要特点\n- 标准 API 价格比 GPT-5.6 Terra 与 GPT-6.1 Sol 低 50%，在 ChatGPT Work 与 Codex 向 Plus、Pro、Enterprise、Edu 用户开放，并可通过 API 使用。[^s1]\n- 在 OpenAI Preparedness Framework 下达到网络安全 Critical 阈值，并沿用为 GPT-6 Astra 开发的防护措施。[^s1]\n\n## 定位与后续\nGPT-6 世代中定位在 GPT-6 Luna 之上、GPT-6.1 Sol 与旗舰 GPT-6 Astra 之下的中端高效模型，是 GPT-5.6 Terra 的直接继任者（价格减半），与 GPT-6.1 Sol 同属 GPT-6.1 同代更新但更便宜、更快，主打视觉推理与电脑操作而非最难的编码任务。 其同家族更强的模型为[[work_gpt_6_1_sol|GPT-6.1 Sol]]。[^s1]",
    "url": "/works/gpt-6-1-terra.html",
    "lang": "ms",
    "sourceLang": "zh",
    "fallback": true,
    "availableLangs": [
      "zh"
    ],
    "workKind": "ai_model"
  },
  "entities": null,
  "relations": [
    {
      "id": "relation_work_gpt_6_1_terra_published_by_openai",
      "fromId": "work_gpt_6_1_terra",
      "relationType": "published_by",
      "toId": "organization_openai",
      "sourceRefs": [
        {
          "entityId": "work_gpt_6_1_terra",
          "sourceId": "s1"
        }
      ],
      "label": "Published by",
      "targetId": "organization_openai",
      "targetTitle": "OpenAI",
      "targetUrl": "/ms/organizations/openai.html",
      "incoming": false
    }
  ],
  "mentions": [
    {
      "fromId": "work_gpt_5_6_terra",
      "fromTitle": "GPT-5.6 Terra",
      "fromUrl": "/ms/works/gpt-5-6-terra.html",
      "fromType": "work"
    }
  ],
  "relationGroups": [
    {
      "type": "published_by",
      "label": "Published by",
      "items": [
        {
          "id": "relation_work_gpt_6_1_terra_published_by_openai",
          "fromId": "work_gpt_6_1_terra",
          "relationType": "published_by",
          "toId": "organization_openai",
          "sourceRefs": [
            {
              "entityId": "work_gpt_6_1_terra",
              "sourceId": "s1"
            }
          ],
          "label": "Published by",
          "targetId": "organization_openai",
          "targetTitle": "OpenAI",
          "targetUrl": "/ms/organizations/openai.html",
          "incoming": false
        }
      ]
    }
  ],
  "lang": "ms",
  "availableLangs": [
    "zh"
  ],
  "fallback": true,
  "infoFields": [
    {
      "key": "work_kind",
      "label": "Work kind",
      "value": "Model AI"
    }
  ],
  "pagination": {
    "page": 1,
    "pageSize": 50,
    "total": 1,
    "pageCount": 1
  },
  "relationsTotal": 1,
  "mentionsTotal": 1
}