All tasks / Models
olmOCR-2-7B-1025
Ai2's open-weights 7B vision-language model and toolkit for converting PDFs and scans to clean Markdown.
- Allen Institute for AI (Ai2)
- Open weights
- Released 21 Oct 2025
Where it ranks
- –Documents & OCRModels · Local · no result on this task’s benchmarks yet
Facts
- Vendor
- Allen Institute for AI (Ai2)
- Family
- olmOCR
- Released
- 21 Oct 2025
- Weights
- Open: download and run it yourself
- License
- Apache-2.0
- Parameters
- 8.29B
- GPU memory to run locally
- ~12 GB (Official: the olmOCR README asks for an NVIDIA GPU with at least 12 GB of GPU RAM (default FP8 checkpoint).)
- API price
- $0.09 in / $0.19 out per 1M tokens (DeepInfra, as listed among verified external providers in the olmOCR README (Cirrascale $0.07/$0.15, Parasail $0.10/$0.20).)
- Access
- Free (open weights)
- Last verified
- 18 Oct 2026
Price from github.com, 18 Oct 2026.
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| MDPBench (multilingual document parsing) | 70.4% | olmOCR2Overall (public split, page-level mean of text, formula and table scores) from the board's leaderboard.json (updated 2026-10-04). | 4 Oct 2026 | huggingface.co |
| OCR Arena | 1,400 | olmOCR 2Elo read from the site's JSON endpoint https://www.ocrarena.ai/api/leaderboard; 2204 battles. | 18 Oct 2026 | ocrarena.ai |
Where you can use it
No tracked tool lists this model yet; as open weights it runs in local runtimes.
Earlier versions
- olmOCR-7B-0825 · 13 Aug 2025
- olmOCR-7B-0725 · 24 Jul 2025