ParseBench
A public evaluation benchmark designed specifically for real-world PDF parsing scenarios. It covers sub-tasks including table extraction, multi-column body text, and scanned document recognition, and measures model performance across multiple dimensions such as precision, recall, and cost per thousand pages.
Timeline (last 90 days)
ParseBench's overall ranking is a weighted aggregate of multiple dimensions, and standout sub-items do not necessarily translate into an overall lead.
ParseBench is a public benchmark platform for document parsing tasks.
GPT 6.1 Terra ranks fifth on the overall ParseBench leaderboard, above 6 Sol and below Opus 5.5 and Astra.
ParseBench published the document parsing evaluation results for GPT 6.1 Terra.
Among document parsing benchmarks, the top of ParseBench is KDLAI/KDL-Frontier-Parser-nano (76.36), olmOCR-bench belongs to infly/Infinity-Parser2-Pro (87.6), and Real5-OmniDocBench is taken by PaddlePaddle/PaddleOCR-VL-1.6 with 93.19.
According to the official ParseBench benchmark, LightOn OCR-3 sits on the Pareto frontier of open-weight OCR models.
LightOn OCR-3 costs approximately $3.19 per thousand pages of documents (based on ParseBench calculations).
OpenDocRouter uses an evaluate-then-onboard mechanism: after a new OCR model is released, it first completes benchmarking on ParseBench and is then integrated into OpenDocRouter.
OpenDocRouter's companion evaluation platform ParseBench has catalogued 130+ models.
ParseBench is a comprehensive evaluation benchmark for document parsing capabilities launched by LlamaIndex, covering dimensions such as OCR text recognition, table structure reconstruction, and reading order.
2 more timeline events
All Facts (10)
GPT 6.1 Terra ranks fifth on the overall ParseBench leaderboard, above 6 Sol and below Opus 5.5 and Astra.
50%UnverifiedParseBench's overall ranking is a weighted aggregate of multiple dimensions, and standout sub-items do not necessarily translate into an overall lead.
50%UnverifiedParseBench is a public benchmark platform for document parsing tasks.
50%UnverifiedAmong document parsing benchmarks, the top of ParseBench is KDLAI/KDL-Frontier-Parser-nano (76.36), olmOCR-bench belongs to infly/Infinity-Parser2-Pro (87.6), and Real5-OmniDocBench is taken by PaddlePaddle/PaddleOCR-VL-1.6 with 93.19.
50%UnverifiedAccording to the official ParseBench benchmark, LightOn OCR-3 sits on the Pareto frontier of open-weight OCR models.
50%UnverifiedOpenDocRouter uses an evaluate-then-onboard mechanism: after a new OCR model is released, it first completes benchmarking on ParseBench and is then integrated into OpenDocRouter.
50%UnverifiedA systematic evaluation covered the performance of 16 frontier vision-language models (VLMs) in document parsing scenarios.
50%UnverifiedParseBench is a publicly available document parsing benchmark designed for real-world PDF parsing scenarios, covering sub-tasks such as table extraction, multi-column body text, and scanned document recognition.
50%UnverifiedParseBench published the document parsing evaluation results for GPT 6.1 Terra.
50%UnverifiedParseBench is a comprehensive evaluation benchmark for document parsing capabilities launched by LlamaIndex, covering dimensions such as OCR text recognition, table structure reconstruction, and reading order.
50%