Unlimited-OCR
Unlimited-OCR is the June 2026 predecessor to Unlimited-OCR-2. It builds on DeepSeek-OCR and targets multi-page parsing in a single document-level pass.
Model design and access
The repository provides downloadable weights and documents Transformers, vLLM and SGLang inference paths. It treats the document as a sequence rather than a collection of independent OCR requests. This entry retains the available checkpoint; it does not claim that the original hosted API remains separately selectable.Unlimited-OCR official model card
Practical use
Practical guidance. Keep it for a reproducible document pipeline or a comparison with the newer release. Include documents where table rows, headings and reading order continue across page boundaries. Verify those links directly instead of scoring only the number of recognized characters.Unlimited-OCR official model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| License | MITUnlimited-OCR official model card |
| Lineage | Built on DeepSeek-OCRUnlimited-OCR official model card |
Related models
- Unlimited-OCR-2
Baidu · Unlimited-OCR
Document parsing model with sparse-reference sliding-window attention for long documents.
Launch date Downloadable · API