OLMOCR-BENCH · OLMOCR-BENCH · CHANDRA 2.5
olmOCR-bench: Chandra 2.5 scores 86.9% on olmOCR-bench · Chandra 2.5.
olmOCR-bench unit-tests OCR output on arXiv papers, old scans, tables, multi-column pages and tiny text. Chandra 2.5 is the model behind the Datalab API. We report each model with its own postprocessing, because benchmark-specific postprocessing flatters scores.
Ranked scoreboard.
Models ranked highest-to-lowest. Datalab variants in accent; competitors and prior generations in muted ink.
Rank Model Score
01 Chandra 2.5 Datalab API 86.9%
02 Chandra OCR 2 open source 85.8%
03 LightOnOCR-3-4B our reproduction 85.4%
+1.1 vs Chandra OCR 2
olmocr 0.4.27 · olmOCR-bench revision 54a96a6 · headers and footers left out
olmOCR-bench unit-tests OCR output against known-correct elements in a public dataset. Model versions pinned per run; anyone can reproduce from the olmOCR-bench HuggingFace dataset.
OTHER BENCHMARKS
All benchmarks →Compare on another doc type.
Overall Every test: text, tables and layout 93.5% 93.5 for Datalab accurate, the top score of every parser we ran. Tables Financial filings · regulatory PDFs · research data 95.2% 95.2 for Datalab accurate on 6,767 table tests, the top score of every parser. Each cell must sit under the right headings and beside the right neighbors.
START
Try our hosted API on your hardest PDF.
Free tier. No credit card. SOC 2 Type II.