{
  "schemaVersion": "1.0",
  "updates": [],
  "companyNews": [],
  "usageGuide": {
    "facts": [
      {
        "key": "limits",
        "label": "使用前看",
        "text": "代码和本次核对的 olmOCR-2-7B-1025-FP8 模型卡均标注 Apache 2.0。README 的本地 GPU 方案要求相应 NVIDIA 环境及显存；远程方案把推理交给所连接服务器，不能描述为全程离线。",
        "sourceUrls": [
          "https://github.com/allenai/olmocr/blob/main/README.md",
          "https://github.com/allenai/olmocr/blob/main/LICENSE",
          "https://huggingface.co/allenai/olmOCR-2-7B-1025-FP8"
        ]
      }
    ]
  },
  "id": "b928",
  "language": "zh",
  "name": "olmOCR",
  "originalName": "olmOCR",
  "library": "tools",
  "category": "模型开发",
  "categoryKey": "模型开发",
  "categoryKeys": [
    "模型开发"
  ],
  "categories": [
    "模型开发"
  ],
  "entryPoints": [],
  "summary": "用视觉语言模型将 PDF 和文档图像转成保留阅读顺序的 Markdown。",
  "introduction": "olmOCR 面向需要批量整理文档内容的开发者和研究者，以视觉语言模型把 PDF、PNG、JPEG 线性化为 Markdown，并处理多栏、表格和公式等阅读顺序问题。工具既可连接远程 vLLM 服务，也可使用本地 GPU 推理；它提供的是处理流水线，不是下载即得的数据集。官网发布的模型卡与代码分别提供使用信息，结果仍需根据原文检查。",
  "capabilities": [],
  "company": {
    "name": "艾伦人工智能研究所",
    "originalName": "The Allen Institute for Artificial Intelligence",
    "country": {
      "codes": [
        "US"
      ],
      "names": [
        "美国"
      ],
      "display": "美国",
      "basis": "company-location",
      "basisLabel": "公开公司地址"
    }
  },
  "officialUrl": "https://github.com/allenai/olmocr",
  "profileUpdatedOn": "2026-10-23",
  "sources": [
    {
      "kind": "product",
      "url": "https://github.com/allenai/olmocr"
    },
    {
      "kind": "documentation",
      "url": "https://github.com/allenai/olmocr/blob/main/README.md"
    },
    {
      "kind": "documentation",
      "url": "https://huggingface.co/allenai/olmOCR-2-7B-1025-FP8"
    }
  ],
  "news": [],
  "links": {
    "interactive": "products/olmocr--b928.html",
    "reading": "read/zh/products/olmocr--b928.html",
    "markdown": "read/zh/products/olmocr--b928.md",
    "json": "read/zh/products/olmocr--b928.json"
  }
}