chandra-OCR-GGUF

Chandra is a highly accurate OCR model designed to convert images and PDFs into structured outputs such as markdown, HTML, and JSON while preserving detailed layout information. It supports over 40 languages and excels in handling complex document elements including handwriting, tables, math expressions, forms with checkboxes, and diagrams with captions. Chandra offers flexible inference modes with local execution via HuggingFace or remote deployment using a vLLM server, making it suitable for both interactive use and large-scale batch processing. Its strong layout preservation and multilingual capabilities make it a versatile choice for document digitization and automated content extraction workflows.

Model Files

File Name	Quant Type	File Size
chandra-BF16.gguf	BF16	16.4 GB
chandra-F16.gguf	F16	16.4 GB
chandra-F32.gguf	F32	32.8 GB
chandra-Q3_K_M.gguf	Q3_K_M	4.12 GB
chandra-Q3_K_S.gguf	Q3_K_S	3.77 GB
chandra-Q4_K_M.gguf	Q4_K_M	5.03 GB
chandra-Q4_K_S.gguf	Q4_K_S	4.8 GB
chandra-Q8_0.gguf	Q8_0	8.71 GB
chandra-mmproj-bf16.gguf	mmproj-bf16	1.16 GB
chandra-mmproj-f16.gguf	mmproj-f16	1.16 GB
chandra-mmproj-f32.gguf	mmproj-f32	2.31 GB
chandra-mmproj-q8_0.gguf	mmproj-q8_0	752 MB

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):