Benchmark

OCRBench_V2

OCRBench v2: Enhanced large-scale bilingual benchmark for evaluating Large Multimodal Models on visual text localization and reasoning with 10,000…

Modality
multimodal
Categories
image_to_text, vision
Openness
unknown
Source
llm_stats
Reported scores
7

Reported scores

llm_stats

ModelOrganizationReported valueReportedEvidence
Nova 2 LiteAmazon0.5612025-12-02evidence
Nova 2 OmniAmazon0.5822025-12-02evidence
Nova 2 ProAmazon0.6452025-12-02evidence
Qwen2.5-Omni-7BQwen0.5782025-03-27evidence
Qwen3.7-PlusQwen0.6712026-05-31evidence
Seed 2.1 ProByteDance0.6322026-06-24evidence
Seed 2.1 TurboByteDance0.6282026-06-24evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.