Developer & Productivity Tools · Paid
DeepSeek OCR API
DeepSeek OCR API is a standalone document-recognition endpoint in the DeepSeek API, released on November 13, 2026 as part of what DeepSeek calls its full-suite app API release. It turns PDFs, Word and PowerPoint files and images into text, page by page, and is one of the DeepSeek App capabilities the company opened up as standalone APIs. The endpoint is POST https://api.deepseek.com/v1/tools/ocr, authenticated with a normal DeepSeek API key.
Who it is for: developers building retrieval-augmented generation (RAG) pipelines, document assistants, invoice or form extraction, search indexing and archive digitization who want raw page text without sending every document through a chat model. Recognized files are not stored in the DeepSeek Files API and are not passed to a chat model. To have a model answer questions about a document instead, DeepSeek offers file input (file parsing) inside its Messages, Responses and Chat Completions APIs on the deepseek-flash model.
Key features: - Multipart upload with two required fields: the file and a purpose. - Two modes: text_recognition, which extracts text directly from Word and PowerPoint files that contain no images and recognizes other files page by page, and rich_layout_recognition, which recognizes every file page by page. - Supported inputs: PDF, DOC, DOCX, PPT, PPTX and common image formats such as PNG, JPEG, WebP, GIF, BMP and TIFF, detected from the file content. - Limits: files up to 100 MiB and 1,000 pages; the upload must finish within 120 seconds and the whole request within 180 seconds. - The JSON response lists every page with its index, the raw model output and a finish_reason that flags pages cut off at the output limit, plus token usage for reference.
Pricing type: paid, pay as you go from a prepaid DeepSeek balance. As of November 14, 2026, file parsing and OCR cost $0.225 per 1,000 pages on DeepSeek's Models and Pricing page, so a 10,000-page archive costs about $2.25 to recognize. DeepSeek says its prices may change.
Related DeepSeek tool APIs launched the same day: Search and Fetch ($0.75 per 1,000 uses each), speech to text (ASR) and text to speech (TTS), which were on limited-time 50% discounts at launch.
Alternatives: Mistral OCR, Google Cloud Document AI, Amazon Textract, Azure AI Document Intelligence, open-source pipelines such as Tesseract or docling, and multimodal chat models that read documents directly, including Gemini and Claude.
Official documentation: https://api-docs.deepseek.com/guides/ocr
