Qianfan OCR vs olmOCR
Pick Qianfan OCR for self-hosted multilingual parsing that needs more than Markdown, such as JSON, HTML, LaTeX or bounding boxes, and for key information extraction. Pick olmOCR for large-scale PDF to Markdown for LLM training or RAG, where its stated cost is under $200 per million pages and it scores 82.4 on olmOCR-Bench. Qianfan OCR is the smaller model at 4B against 7B, but olmOCR has the clearer hardware floor of a 12 GB GPU.
Qianfan OCRbaidu/qianfan-ocr | olmOCRallenai/olmocr | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 1.5 s / page |
| Best for | Chinese documents | Bulk PDF to text for datasets |
| Inputs | PDF, images | PDF, images |
| Languages | Chinese, English and more | Mainly English |
| Max pages per request | 200 | 1,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | No |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose Qianfan OCR if
- It covers 192 languages, handwriting and scene text, which suits multilingual document parsing.
- It outputs Markdown, JSON, HTML, LaTeX and bounding boxes, while olmOCR lists only Markdown and JSON.
- It scores 93.12 on OmniDocBench v1.5 and is also offered hosted on the Baidu Qianfan platform.
Choose olmOCR if
- It scores 82.4 on olmOCR-Bench, above Marker (76.1), MinerU 2.5.4 (75.2) and Mistral OCR (72.0), and above the 79.8 that Qianfan OCR reports on the same benchmark.
- Ai2 puts self-hosted cost under $200 per million pages, and hosted olmOCR-2-7B on DeepInfra costs $0.09 input and $0.19 output per million tokens.
- It handles equations, tables and handwriting, and strips headers and footers in natural reading order, which fits academic papers and mixed printed and handwritten pages.
Key differences
| Qianfan OCR | olmOCR | |
|---|---|---|
| Model size | 4B (Qwen3-4B language model) | 7B |
| Licence | Apache-2.0 | Apache-2.0 (toolkit) |
| Output formats | Markdown, JSON, HTML, LaTeX, bounding boxes | Markdown, JSON |
| Hardware | Benchmarked on one NVIDIA A100 with vLLM; no minimum VRAM stated | NVIDIA GPU with at least 12 GB VRAM and 30 GB free disk |
| Pricing | Free self-hosted (you pay for compute); hosted via Baidu Qianfan platform | Self-hosted under $200 per million pages; DeepInfra $0.09 input / $0.19 output per million tokens |
| olmOCR-Bench score | 79.8 (reported by the authors) | 82.4 (v0.4.0, reported by the authors) |
| Known failure | Thinking mode can emit up to 16,384 tokens, adding latency; 1.024 pages per second on an A100 even with W8A8 | Out of memory on a 12 GB GPU (issue #424), gets stuck on the last few pages (issue #433), unreliable image bounding box coordinates (issue #437) |
Sources
- Qianfan-OCR model card
- Qianfan-OCR paper
- olmOCR GitHub README
- olmOCR-2 model
- Issue #424 OOM
- Issue #437 bbox
What each one is
Qianfan OCR
Baidu's OCR-tuned vision model. Strong on Chinese documents and tables.
Qianfan OCR detailsolmOCR
Open-source vision model from Ai2 (often searched as olm ocr) that turns PDFs into clean, reading-order text. Latest release: olmOCR 2.
olmOCR detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "baidu/qianfan-ocr","fallbacks": ["allenai/olmocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Qianfan OCR alternativesolmOCR alternativesBest OCR models in 2026