Traditional OCR pipelines chain at least three models: a layout detector, a text recognizer, and a language model for understanding. Qianfan-OCR replaces all three with a single 4B model that goes directly from image to Markdown.
The key innovation is Layout-as-Thought: appending a <think> token to any prompt
triggers an optional reasoning phase where the model explicitly works through bounding
boxes, element types, and reading order before producing output. It’s Chain-of-Thought
for document layout - and it’s optional, so you can skip it for simple single-column
documents to save latency.
The benchmark that caught my attention: two-stage OCR+LLM pipelines score 0.0 on CharXiv chart reasoning benchmarks. When a pipeline discards visual structure during the OCR step, the downstream LLM has nothing to reason over. End-to-end models don’t have this problem.
Results: #1 on OmniDocBench v1.5 (93.12, end-to-end), #1 on OCRBench (880 overall), #1 on KIE (87.9, beating Gemini-3.1-Pro and Qwen3-VL-235B). Runs at 1.024 pages/sec on a single A100 with W8A8 quantization.