olmOCR 2 7B 1025
olmOCR 2 converts document images into structured text.
What the release provides
It adapts Qwen2.5-VL-7B-Instruct with OCR data and GRPO training for equations and tables. Ai2 recommends the FP8 release for inference and BF16 for further fine-tuning. The toolkit supplies page rendering and metadata. Apache 2.0 applies.olmOCR 2 7B 1025 official model card
How to use it well
Practical guidance. Review reading order, equations and table structure against the page image. Keep a trace from each extracted passage to its document page so users can check errors.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | allenai/olmOCR-2-7B-1025 |