DeepSeek-OCR

DeepSeek-OCR is a comprehensive Optical Character Recognition (OCR) model that analyzes and understands complex documents. It excels at challenging OCR tasks, including recognizing mathematical formulas and processing text that is curved, rotated, or overlapping.

View model card in Model Garden

Supported regions
Model ID	`deepseek-ocr-maas`
Launch stage	GA
Supported inputs & outputs	Inputs: Text, Documents, Images Outputs: Text
Capabilities	Supported Not supported Batch predictions Function calling Structured output
Usage types	Supported Dynamic shared quota Not supported Fixed quota Provisioned Throughput
Versions	`DeepSeek-OCR` Launch stage: GA Release date: October 23, 2025
	Model availability	United States `us-central1`
	ML processing	United States `Multi-region`
Limits	us-central1: Max output: 8192 Context length: 8192
Pricing	See Pricing.

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2025-11-06 UTC.

DeepSeek-OCR Stay organized with collections Save and categorize content based on your preferences.

DeepSeek-OCR