multimodal benchmark
CC-OCR
A comprehensive OCR benchmark for evaluating Large Multimodal Models (LMMs) in literacy. Comprises four OCR-centric tracks: multi-scene text reading, multilingual text reading, document parsing, and key information extraction. Contains 39 subsets with 7,058 fully annotated images, 41% sourced from real applications. Tests capabilities including text grounding, multi-orientation text recognition, and detecting hallucination/repetition across diverse visual challenges.
Updated Aug 7, 2026
Published models18
Registry coverage18
MetricScore
EvidenceB
CC-OCR leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
Score distribution
Top published rows on the benchmark's original scale.
CC-OCR highlights
The top published results on this benchmark's own scale.
What is CC-OCR?
Definition and scoring fields from the benchmark registry.
A comprehensive OCR benchmark for evaluating Large Multimodal Models (LMMs) in literacy. Comprises four OCR-centric tracks: multi-scene text reading, multilingual text reading, document parsing, and key information extraction. Contains 39 subsets with 7,058 fully annotated images, 41% sourced from real applications. Tests capabilities including text grounding, multi-orientation text recognition, and detecting hallucination/repetition across diverse visual challenges.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- CC-OCR
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- cc-ocr|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about CC-OCR.
Which model scores highest on CC-OCR?
Qwen3.6 Plus is currently ranked first with 83.4%.
What does CC-OCR measure?
A comprehensive OCR benchmark for evaluating Large Multimodal Models (LMMs) in literacy. Comprises four OCR-centric tracks: multi-scene text reading, multilingual text reading, document parsing, and key information extraction. Contains 39 subsets with 7,058 fully annotated images, 41% sourced from real applications. Tests capabilities including text grounding, multi-orientation text recognition, and detecting hallucination/repetition across diverse visual challenges.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
18 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.