What are the hardware requirements for PaddleOCR VL?

PaddleOCR VL requires approximately 1.9 GB of VRAM and supports a context window of 128K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed PaddleOCR VL and how large is it?

PaddleOCR VL is developed by unsloth, a model with 0.3b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate PaddleOCR VL?

PaddleOCR VL is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

PaddleOCR VL by unsloth — VRAM 1.9GB, 128K context

Name: PaddleOCR VL
Author: unsloth

PaddleOCR VL is an open-source language model by unsloth. Features: 0.3b LLM, VRAM: 1.9GB, Context: 128K, License: apache-2.0, LLM Explorer Score: 0.22.

Arxiv:2510.14528 Base model:baidu/ernie-4.5-0.3... Base model:finetune:baidu/erni... Chart Conversational Custom code Document-parse En Ernie4.5 Formula Image-text-to-text Image-to-text Layout Multilingual Ocr Paddleocr Paddleocr vl Paddlepaddle Region:us Safetensors Table Zh

Model Card on HF 🤗: https://huggingface.co/unsloth/PaddleOCR-VL

PaddleOCR VL Parameters and Internals

LLM Name	PaddleOCR VL
Repository 🤗	https://huggingface.co/unsloth/PaddleOCR-VL
Base Model(s)	baidu/ERNIE-4.5-0.3B-Paddle baidu/ERNIE-4.5-0.3B-Paddle
Model Size	0.3b
Required VRAM	1.9 GB
Updated	2026-08-10
Maintainer	unsloth
Model Type	paddleocr_vl
Model Files	1.9 GB
Supported Languages	en zh
Model Architecture	PaddleOCRVLForConditionalGeneration
License	apache-2.0
Context Length	131072
Model Max Length	131072
Transformers Version	4.55.0
Tokenizer Class	LlamaTokenizer
Padding Token	<unk>
Vocabulary Size	103424
Torch Data Type	bfloat16

Rank the PaddleOCR VL capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to PaddleOCR VL

Best Alternatives	Context / RAM	Downloads	Likes
PaddleOCR VL 1.5 Bf16	128K / 1.8 GB	639	6
PaddleOCR VL Bfloat16	128K / 1.8 GB	44	1
PaddleOCR VL 1.5 8bit	128K / 1.1 GB	137	1
PaddleOCR VL 1.5 4bit	128K / 0.7 GB	105	3
PaddleOCR VL 8bit	128K / 1.1 GB	52	3
PaddleOCR VL 4bit	128K / 0.7 GB	49	2

Note: green Score (e.g. "73.2") means that the model is better than unsloth/PaddleOCR-VL.