Qianfan-OCR: A Unified End-to-End Model for Document Intelligence

TL;DR

Qianfan-OCR is a large unified vision-language model that integrates document parsing, layout analysis, and understanding, enabling direct image-to-Markdown conversion and supporting diverse prompt-driven document tasks.

Contribution

It introduces Layout-as-Thought, a novel thinking phase that recovers layout grounding in end-to-end OCR models, enhancing accuracy on complex layouts.

Findings

Ranks first on OmniDocBench v1.5 with 93.12 score

Achieves top results on OlmOCR Bench with 79.8 score

Outperforms several models on key information extraction benchmarks

Abstract

We present Qianfan-OCR, a 4B-parameter end-to-end vision-language model that unifies document parsing, layout analysis, and document understanding within a single architecture. It performs direct image-to-Markdown conversion and supports diverse prompt-driven tasks including table extraction, chart understanding, document QA, and key information extraction. To address the loss of explicit layout analysis in end-to-end OCR, we propose Layout-as-Thought, an optional thinking phase triggered by special think tokens that generates structured layout representations -- bounding boxes, element types, and reading order -- before producing final outputs, recovering layout grounding capabilities while improving accuracy on complex layouts. Qianfan-OCR ranks first among end-to-end models on OmniDocBench v1.5 (93.12) and OlmOCR Bench (79.8), achieves competitive results on OCRBench, CCOCR, DocVQA, and ChartQA against general VLMs of comparable scale, and attains the highest average score on public key information extraction benchmarks, surpassing Gemini-3.1-Pro, Seed-2.0, and Qwen3-VL-235B. The model is publicly accessible via the Baidu AI Cloud Qianfan platform.