What are the hardware requirements for Irwan LightOnOCR 3 0.8B Abliterated?

Irwan LightOnOCR 3 0.8B Abliterated requires approximately 0.2 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Irwan LightOnOCR 3 0.8B Abliterated and how large is it?

Irwan LightOnOCR 3 0.8B Abliterated is developed by irwanworld, a model with 0.8b 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 Irwan LightOnOCR 3 0.8B Abliterated?

Irwan LightOnOCR 3 0.8B Abliterated is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Irwan LightOnOCR 3 0.8B Abliterated by irwanworld — VRAM 0.2GB

Name: Irwan LightOnOCR 3 0.8B Abliterated
Author: irwanworld

Irwan LightOnOCR 3 0.8B Abliterated is an open-source language model by irwanworld. Features: 0.8b LLM, VRAM: 0.2GB, License: apache-2.0, Quantized, LLM Explorer Score: 0.3.

Arxiv:2601.14251 Abliterated Base model:huihui-ai/huihui-qw... Base model:quantized:huihui-ai... Conversational Document-parsing En Endpoints compatible Gguf Id Image-text-to-text Llama.cpp Ocr Q8 Quantized Qwen3 5 Region:us Safetensors Uncensored Vision

Model Card on HF 🤗: https://huggingface.co/irwanworld/Irwan-LightOnOCR-3-0.8B-abliterated

Irwan LightOnOCR 3 0.8B Abliterated Parameters and Internals

LLM Name	Irwan LightOnOCR 3 0.8B Abliterated
Repository 🤗	https://huggingface.co/irwanworld/Irwan-LightOnOCR-3-0.8B-abliterated
Base Model(s)	lightonai/LightOnOCR-3-0.8B huihui-ai/Huihui-Qwen3.5-0.8B-abliterated lightonai/LightOnOCR-3-0.8B huihui-ai/Huihui-Qwen3.5-0.8B-abliterated
Model Size	0.8b
Required VRAM	0.2 GB
Updated	2026-10-21
Maintainer	irwanworld
Model Type	qwen3_5
Model Files	0.8 GB 0.2 GB 1.7 GB
Supported Languages	en id
GGUF Quantization	Yes
Quantization Type	q8\|gguf
Model Architecture	Qwen3_5ForConditionalGeneration
License	apache-2.0
Model Max Length	262144
Transformers Version	5.5.4
Tokenizer Class	TokenizersBackend
Padding Token	<\|endoftext\|>
Errors	replace

Rank the Irwan LightOnOCR 3 0.8B Abliterated 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 Irwan LightOnOCR 3 0.8B Abliterated

Best Alternatives	Context / RAM	Downloads	Likes
...5 0.8B GPT6 Luna Distill Karpy	0K / 0.2 GB	884	1
Loopl 0.8B 4bit	0K / 0.6 GB	146	1
Mavery Lehrling 0.8B MLX 4bit	0K / 0.4 GB	77	1
Qwen3.5 0.8B MLX 4bit	0K / 0.6 GB	5113	5
Qwen3.5 0.8B 8bit	0K / 1 GB	4792	2
Qwen3.5 0.8B 4bit	0K / 0.6 GB	2549	3
Qwen3.5 0.8B OptiQ 4bit	0K / 0.2 GB	1494	22
Chartreader 0.8B OptiQ 4bit	0K / 0.7 GB	105	2
OvisOCR2	0K / 1.7 GB	184769	498
Qwen3.5 0.8B	0K / 1.8 GB	2839884	655

Note: green Score (e.g. "73.2") means that the model is better than irwanworld/Irwan-LightOnOCR-3-0.8B-abliterated.