cstr/paddleocr-vl-0.9b-GGUF

PaddleOCR-VL-0.9B — CrispEmbed GGUF

CrispEmbed-native GGUF quantizations of PaddlePaddle/PaddleOCR-VL.

End-to-end VLM-based OCR: text recognition, table extraction, formula recognition, chart understanding. 109 languages.

Files

Quantization quality. This is a small 0.9B model, so 4-bit quantization is noticeably lossy. In testing, q8_0 transcribes fully and matches the fp16 reference, while q4_k tends to terminate early (e.g. it emitted only "The quick brown fox jumps" and stopped, dropping the rest of the line) — a flattened output distribution from 4-bit weights, not a bug. Behavior is identical on CPU and Metal. Use q8_0 for accurate OCR; reach for q4_k only when memory is tight and truncated/approximate output is acceptable.

Model

• —Architecture: NaViT-style ViT (27L, 1152d, SigLIP 2D RoPE + learned position embeddings)
• —Projector (pre-norm → 2×2 spatial merge → MLP)
• —ERNIE-4.5-0.3B LLM decoder (18L, 1024d, 16/2 GQA, MRoPE, SwiGLU)
• —Parameters: ~0.9B total
• —Languages: 109 (multilingual)
• —Tasks: OCR, Table Recognition, Formula Recognition, Chart Recognition
• —License: Apache 2.0

Usage with CrispEmbed

# OCR
./crispembed -m paddleocr-vl-0.9b-q8_0.gguf --ocr document.png

# With specific prompt
./crispembed -m paddleocr-vl-0.9b-q8_0.gguf --ocr-prompt "Table Recognition:" table.png

Conversion

git clone https://github.com/CrispStrobe/CrispEmbed
cd CrispEmbed

python models/convert-paddleocr-vl-to-gguf.py \\
    --model PaddlePaddle/PaddleOCR-VL \\
    --output paddleocr-vl-0.9b-f16.gguf --dtype f16

./build/crispembed-quantize paddleocr-vl-0.9b-f16.gguf paddleocr-vl-0.9b-q8_0.gguf q8_0

License

Apache 2.0 — same as the base model.

Provenance and EU AI Act Art. 53 note

• —Upstream model: PaddlePaddle/PaddleOCR-VL — published by PaddlePaddle.
• —Upstream licence: apache-2.0. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• —What was done here: format conversion and/or quantisation only (GGUF). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• —Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository. No training-content summary was found on the upstream model card at the time of writing; that documentation gap is upstream's and is not filled here.
• —Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.