AI OCR models
103 OCR APIs, vision models and open-source engines. Same request, same output, one bill.
| Model | Best for | Supports | Speed | Price |
|---|---|---|---|---|
Tesseracttesseract/tesseract | Clean printed text, free | Multilingual | 0.6 s / page | Free |
DeepSeek-OCRdeepseek/deepseek-ocr | Cheap bulk conversion | TablesMathBounding boxesMultilingual | 1.1 s / page | 1 credit / page |
Jina OCRjina/jina-ocr | Clean Markdown at low cost | TablesMathMultilingual | 1.5 s / page | 1 credit / page |
PaddleOCR (PP-OCR)paddle/paddleocr | Fast printed text, Chinese and English | Bounding boxesConfidenceMultilingual | 0.7 s / page | 1 credit / page |
EasyOCRjaided/easyocr | Quick scripts and scene text | Bounding boxesConfidenceMultilingual | 0.9 s / page | 1 credit / page |
OCRmyPDFocrmypdf/ocrmypdf | Searchable PDFs from scans | Multilingual | 1.0 s / page | 1 credit / page |
RapidOCRrapidai/rapidocr | Lightweight OCR on CPU or edge | Bounding boxesConfidenceMultilingual | 0.6 s / page | 1 credit / page |
docTRmindee/doctr | Word-level OCR with positions | Bounding boxesConfidence | 0.9 s / page | 1 credit / page |
Krakenkraken/kraken | Historical documents and manuscripts | HandwritingBounding boxesMultilingual | 1.5 s / page | 1 credit / page |
TrOCRmicrosoft/trocr | Handwritten text lines | Handwriting | 1.2 s / page | 1 credit / page |
Granite Doclingibm/granite-docling | Small-model document conversion | TablesMathBounding boxes | 1.5 s / page | 1 credit / page |
SmolDoclinghuggingface/smoldocling | Very cheap page conversion | TablesMathBounding boxes | 1.0 s / page | 1 credit / page |
Donutnaver/donut | Receipt and form fields after fine-tuning | Forms | 1.0 s / page | 1 credit / page |
GOT-OCR 2.0stepfun/got-ocr-2 | Formatted text, formulas and charts | TablesMathBounding boxes | 1.5 s / page | 1 credit / page |
Florence-2microsoft/florence-2 | Lightweight OCR with boxes | Bounding boxes | 0.8 s / page | 1 credit / page |
Tesseract.jstesseract/tesseract-js | OCR in the browser | Bounding boxesConfidenceMultilingual | 2.0 s / page | 1 credit / page |
OpenOCRfudan/openocr | Lightweight OCR pipeline on CPU | TablesMathBounding boxes | 0.5 s / page | 1 credit / page |
OCR.spaceocrspace/ocr-space | Cheap simple OCR | TablesBounding boxesMultilingual | 1.5 s / page | 1 credit / page |
Gemini 2.5 Flash-Litegoogle/gemini-2-5-flash-lite | High-volume simple pages | TablesMathMultilingual | 2.5 s / page | 1 credit / page |
Qwen2.5-VL-7Bqwen/qwen2-5-vl-7b | Low-cost self-hosted OCR | TablesHandwritingMathBounding boxes | 2.5 s / page | 1 credit / page |
Llama 3.2 11B Visionmeta/llama-3-2-11b-vision | Lightweight self-hosted vision | Multilingual | 3.0 s / page | 1 credit / page |
Mistral Small 3.2mistral/mistral-small-3-2 | Self-hosted small vision model | Multilingual | 3.0 s / page | 1 credit / page |
Amazon Nova Liteamazon/nova-lite | Cheap vision on AWS Bedrock | Multilingual | 3.0 s / page | 1 credit / page |
Phi-3.5-visionmicrosoft/phi-3-5-vision | Small self-hosted model | Multilingual | 3.0 s / page | 1 credit / page |
Kimi-VL-A3Bmoonshot/kimi-vl | Efficient self-hosted vision | Multilingual | 3.0 s / page | 1 credit / page |
GLM-4.1V-Thinkingzai/glm-4-1v-thinking | Small reasoning vision model | Multilingual | 3.0 s / page | 1 credit / page |
SmolVLMhuggingface/smolvlm | On-device and edge use | Multilingual | 3.0 s / page | 1 credit / page |
Idefics3huggingface/idefics3 | Self-hosted image questions | Multilingual | 3.0 s / page | 1 credit / page |
Molmoallenai/molmo | Open research vision model | Multilingual | 3.0 s / page | 1 credit / page |
Llama Nemotron Nano VLnvidia/llama-nemotron-nano-vl | Self-hosted document understanding | TablesMathMultilingual | 2.5 s / page | 1 credit / page |
LLaVA-OneVisionllava/llava-onevision | Open research vision model | Multilingual | 3.0 s / page | 1 credit / page |
Aya Visioncohere/aya-vision | Multilingual image questions | Multilingual | 3.0 s / page | 1 credit / page |
Janus-Prodeepseek/janus-pro | Experimental image understanding | Multilingual | 3.0 s / page | 1 credit / page |
Qianfan OCRbaidu/qianfan-ocr | Chinese documents | TablesMathHandwritingMultilingual | 2.0 s / page | 2 credits / page |
Unlimited OCRbaidu/unlimited-ocr | Books and very long documents | TablesMathBounding boxesMultilingual | 1.2 s / page | 2 credits / page |
MinerUopendatalab/mineru | Academic PDFs | TablesMathBounding boxesMultilingual | 2.0 s / page | 2 credits / page |
Doclingibm/docling | Structured export for RAG | TablesMathBounding boxes | 2.2 s / page | 2 credits / page |
olmOCRallenai/olmocr | Bulk PDF to text for datasets | TablesMathHandwriting | 1.5 s / page | 2 credits / page |
Chandradatalab/chandra | Complex layouts and handwriting | TablesMathHandwritingForms | 2.0 s / page | 2 credits / page |
Suryadatalab/surya | Layout and line-level OCR | TablesBounding boxesMultilingual | 1.5 s / page | 2 credits / page |
dots.ocrrednote/dots-ocr | Multilingual document parsing | TablesMathBounding boxesMultilingual | 1.5 s / page | 2 credits / page |
Nanonets-OCRnanonets/nanonets-ocr | Markdown with semantic tags | TablesMathFormsMultilingual | 1.8 s / page | 2 credits / page |
Hunyuan OCRtencent/hunyuan-ocr | Chinese documents and photos | TablesMathBounding boxesMultilingual | 1.5 s / page | 2 credits / page |
Qwen-VLqwen/qwen-vl | Flexible extraction with prompts | TablesHandwritingMathBounding boxes | 2.0 s / page | 2 credits / page |
MonkeyOCRmonkeyocr/monkeyocr | Chinese and English PDFs | TablesMathBounding boxes | 1.5 s / page | 2 credits / page |
Nougatmeta/nougat | arXiv-style papers | TablesMath | 3.0 s / page | 2 credits / page |
Typhoon OCRtyphoon/typhoon-ocr | Thai documents | TablesMultilingual | 1.8 s / page | 2 credits / page |
RolmOCRreducto/rolmocr | Fast open PDF to text | TablesHandwriting | 3.0 s / page | 2 credits / page |
Pixtralmistral/pixtral | Questions about document images | Multilingual | 3.0 s / page | 2 credits / page |
InternVL3.5opengvlab/internvl | Self-hosted document understanding | TablesMathMultilingual | 3.0 s / page | 2 credits / page |
MiniCPM-V 4.5openbmb/minicpm-v | OCR on edge devices | TablesHandwritingMultilingual | 2.5 s / page | 2 credits / page |
DeepSeek-VL2deepseek/deepseek-vl2 | Charts and tables | TablesBounding boxes | 3.0 s / page | 2 credits / page |
Phi-4-multimodalmicrosoft/phi-4-multimodal | Small model for OCR and questions | TablesMath | 2.5 s / page | 2 credits / page |
Google Cloud Vision OCRgoogle/cloud-vision | Photos and simple scans | HandwritingBounding boxesConfidenceMultilingual | 1.0 s / page | 2 credits / page |
Gemini 2.0 Flashgoogle/gemini-2-0-flash | Low-cost general OCR | TablesHandwritingMathMultilingual | 2.5 s / page | 2 credits / page |
Gemma 3google/gemma-3 | Self-hosted image reading | Multilingual | 3.0 s / page | 2 credits / page |
Qwen3-VLqwen/qwen3-vl | Self-hosted multilingual OCR | TablesHandwritingMathBounding boxes | 2.5 s / page | 2 credits / page |
Llama 4 Maverickmeta/llama-4-maverick | Self-hosted general vision | TablesMathMultilingual | 2.5 s / page | 2 credits / page |
Llama 4 Scoutmeta/llama-4-scout | Self-hosted long inputs | TablesMathMultilingual | 2.5 s / page | 2 credits / page |
ERNIE 4.5 VLbaidu/ernie-4-5-vl | Chinese documents, self-hosted | TablesMathMultilingual | 2.5 s / page | 2 credits / page |
GPT-4.1 miniopenai/gpt-4-1-mini | Low-cost image extraction | TablesMathMultilingual | 2.5 s / page | 3 credits / page |
GPT-4o miniopenai/gpt-4o-mini | Cheap simple pages | TablesMathMultilingual | 2.5 s / page | 3 credits / page |
Gemini 2.5 Flashgoogle/gemini-2-5-flash | Bulk pages with mixed layouts | TablesHandwritingMathMultilingual | 2.5 s / page | 3 credits / page |
Qwen2.5-VL-72Bqwen/qwen2-5-vl-72b | Self-hosted document parsing | TablesHandwritingMathBounding boxes | 2.5 s / page | 3 credits / page |
QVQ-72B-Previewqwen/qvq-72b | Visual reasoning over charts | Multilingual | 3.0 s / page | 3 credits / page |
Llama 3.2 90B Visionmeta/llama-3-2-90b-vision | Self-hosted image questions | Multilingual | 3.0 s / page | 3 credits / page |
Amazon Nova Proamazon/nova-pro | Vision on AWS Bedrock | TablesMathMultilingual | 2.5 s / page | 3 credits / page |
GLM-4.5Vzai/glm-4-5v | Self-hosted document and chart reading | TablesHandwritingMathBounding boxes | 2.5 s / page | 3 credits / page |
Doubao-Seed-1.6bytedance/doubao-seed-1-6 | Chinese documents via Volcano Engine | TablesMathMultilingual | 2.5 s / page | 3 credits / page |
Reka Flashreka/reka-flash | Hosted multimodal questions | Multilingual | 3.0 s / page | 3 credits / page |
GPT-5 miniopenai/gpt-5-mini | Everyday extraction at lower cost | TablesHandwritingMathMultilingual | 2.5 s / page | 4 credits / page |
Qwen-VL-Maxqwen/qwen-vl-max | Hosted Chinese and English documents | TablesHandwritingMathMultilingual | 2.5 s / page | 4 credits / page |
Mistral Medium 3mistral/mistral-medium-3 | Hosted general vision in the EU | TablesMathMultilingual | 2.5 s / page | 4 credits / page |
Command A Visioncohere/command-a-vision | Enterprise document questions | TablesMathMultilingual | 2.5 s / page | 4 credits / page |
LlamaParsellamaindex/llamaparse | RAG ingestion | TablesMathBounding boxesMultilingual | 3.0 s / page | 5 credits / page |
PaddleOCR-VLpaddle/paddleocr-vl | CJK documents | TablesMathBounding boxesConfidence | 1.3 s / page | 6 credits / page |
GLM-OCRzai/glm-ocr | Low-cost general documents | TablesMathBounding boxesConfidence | 1.3 s / page | 6 credits / page |
o4-miniopenai/o4-mini | Reasoning over images at lower cost | TablesHandwritingMathMultilingual | 2.5 s / page | 6 credits / page |
Claude Haiku 4.5anthropic/claude-haiku-4-5 | Fast reading at lower cost | TablesHandwritingMathMultilingual | 2.5 s / page | 6 credits / page |
Grok 2 Visionxai/grok-2-vision | Image questions | Multilingual | 3.0 s / page | 6 credits / page |
Mathpix Convertmathpix/convert | Math and scientific papers | MathHandwritingTablesBounding boxes | 1.5 s / page | 7 credits / page |
Gemini 2.5 Progoogle/gemini-2-5-pro | Long PDFs and complex layouts | TablesHandwritingMathMultilingual | 2.5 s / page | 8 credits / page |
Google Document AIgoogle/document-ai | Multilingual scans | TablesFormsHandwritingBounding boxes | 1.4 s / page | 10 credits / page |
GPTopenai/gpt | Messy scans, custom instructions | TablesHandwritingMathMultilingual | 2.4 s / page | 11 credits / page |
GPT-4.1openai/gpt-4-1 | Long documents with instructions | TablesHandwritingMathMultilingual | 2.5 s / page | 11 credits / page |
GPT-4oopenai/gpt-4o | General document reading | TablesHandwritingMathMultilingual | 2.5 s / page | 11 credits / page |
Markerdatalab/marker | Books and research papers | TablesMathBounding boxesMultilingual | 2.5 s / page | 14 credits / page |
Chunkrchunkr/chunkr | Layout-aware chunks for RAG | TablesMathBounding boxesMultilingual | 3.0 s / page | 14 credits / page |
ABBYY FineReaderabbyy/finereader | Office scans to Word and PDF | TablesBounding boxesMultilingual | 2.0 s / page | 14 credits / page |
Adobe Acrobat OCRadobe/acrobat-ocr | Searchable PDFs in Acrobat | Multilingual | 2.0 s / page | 14 credits / page |
GPT-5openai/gpt-5 | Hard pages with custom instructions | TablesHandwritingMathMultilingual | 2.5 s / page | 14 credits / page |
o3openai/o3 | Charts and pages that need reasoning | TablesHandwritingMathMultilingual | 2.5 s / page | 14 credits / page |
Claude Sonnet 4.5anthropic/claude-sonnet-4-5 | Long dense documents | TablesHandwritingMathMultilingual | 2.5 s / page | 14 credits / page |
Claude Sonnet 4anthropic/claude-sonnet-4 | Dense documents | TablesHandwritingMathMultilingual | 2.5 s / page | 14 credits / page |
Grok 4xai/grok-4 | General image questions | TablesMathMultilingual | 2.5 s / page | 14 credits / page |
Gemini Flashgoogle/gemini-flash | Handwriting and mixed layouts | TablesHandwritingMathMultilingual | 1.6 s / page | 15 credits / page |
Reducto Parsereducto/parse | High-stakes complex documents | TablesFormsBounding boxesConfidence | 4.0 s / page | 20 credits / page |
Unstructuredunstructured/unstructured | Mixed file types into RAG | TablesBounding boxesMultilingual | 3.0 s / page | 20 credits / page |
Mistral OCRmistral/mistral-ocr | High-volume PDF to Markdown | TablesMathBounding boxesMultilingual | 0.9 s / page | 25 credits / page |
Claudeanthropic/claude | Long dense documents | TablesHandwritingMathMultilingual | 2.6 s / page | 25 credits / page |
Amazon Textractaws/textract | Forms and key-value extraction | TablesFormsHandwritingBounding boxes | 1.2 s / page | 25 credits / page |
Claude Opus 4.1anthropic/claude-opus-4-1 | Complex, high-stakes documents | TablesHandwritingMathMultilingual | 2.5 s / page | 27 credits / page |
Azure Document Intelligenceazure/document-intelligence | Enterprise forms and IDs | TablesFormsHandwritingBounding boxes | 1.5 s / page | 60 credits / page |
103 of 103 models