GOT-OCR 2.0
GOT-OCR 2.0 is a 580M-parameter OCR model from joint research involving StepFun, UCAS, Tsinghua University and Megvii. Its compact encoder and decoder turn document images into text.
Extract markup before rendering
The card demonstrates plain text, Markdown and LaTeX output. It also supports region selection by box coordinates or color, and a multi-page mode that produces continuous text. Rendering a formula or chart requires a separate tool.StepFun GOT-OCR-2.0-hf model card
Check reading order and meaning
Practical guidance. Use a representative document with columns, tables and equations. Correct-looking markup can still contain a wrong symbol or reading order. Keep the page image beside extracted text when the result will feed a search index or a later reasoning task.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | GOT-OCR2_0GOT-OCR-2.0-hf |
| Architecture | About 80M parameters in the vision encoder and 0.5B in the decoder.GOT-OCR 2.0 research paper and author affiliations |
| License | Apache 2.0 is declared for the Transformers release.StepFun GOT-OCR-2.0-hf model card |
| Packaging | GOT-OCR2_0 is the original release; GOT-OCR-2.0-hf adapts it for Transformers.StepFun GOT-OCR-2.0-hf model card |
Related models
- Step3-VL-10B
StepFun ยท Step 3
Step3-VL-10B combines a 1.8B PE-lang visual encoder with a Qwen3-8B decoder.
Launch date not confirmedDownloadable