IR

Irwan LightOnOCR 3 0.8B Abliterated

IrwanGGUF by mradermacherUpdated 2026-10-17
Download on Hugging Face
Parameters752M
Context256K
Layers24
Architectureqwen35
Smallest quant0.4 GB
Quick answer

Irwan LightOnOCR 3 0.8B Abliterated needs about 1.4 GB of memory at Q4_K_M (a 0.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 12 quants, from 0.4 GB (Q2_K) to 1.5 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.6 GB.

Irwan LightOnOCR 3 0.8B Abliterated – OCR vision‑language model

Irwan LightOnOCR 3 0.8B Abliterated is a 0.8B OCR vision‑language model based on Qwen3.5. It removes the base model’s rejection bias while retaining OCR capabilities. The model is available in multiple GGUF quantizations.

Good at

  • OCR on images
  • Vision‑language reasoning
  • Tool usage

Limits

  • No training required
  • No vision support beyond OCR
  • License is Apache‑2.0

Best for

  • Transcribing scanned documents
  • Extracting text from images
  • OCR in low‑resource environments

Who it’s for. Developers needing a lightweight OCR model that can run on CPU or GPU with limited VRAM. The model is suitable for integration into local inference pipelines using llama.cpp or transformers.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K0.4 GB 1.3 GB
Fits Get file →
Q3_K_S0.4 GB 1.3 GB
Fits Get file →
Q3_K_M0.5 GB 1.4 GB
Fits Get file →
Q3_K_L0.5 GB 1.4 GB
Fits Get file →
Q4_K_S0.5 GB 1.4 GB
Fits Get file →
IQ4_XS0.5 GB 1.4 GB
Fits Get file →
Q4_K_M0.5 GB 1.4 GB
Fits Get file →
Q5_K_S0.6 GB 1.5 GB
Fits Get file →
Q5_K_M0.6 GB 1.5 GB
Fits Get file →
Q6_K0.6 GB 1.5 GB
Fits Get file →
Q8_00.8 GB 1.7 GB
Fits Get file →
F161.5 GB 2.4 GB
Fits Get file →

For 8 GB VRAM, use Q4_K_S or Q4_K_M (≈1.4 GB memory at 8 k context). If 4 GB VRAM is available, Q2_K (≈1.3 GB) is recommended. For maximum precision, Q8_0 (≈1.7 GB) can be used on systems with 16 GB VRAM.

What GPU do you need for Irwan LightOnOCR 3 0.8B Abliterated?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is F16 (1.5 GB file, about 2.4 GB in use).

Which Irwan LightOnOCR 3 0.8B Abliterated quant should you choose?

Q4_K_M is the usual starting point for Irwan LightOnOCR 3 0.8B Abliterated: it balances file size and quality well. Here it is a 0.5 GB file.

The Q8_0 file is 0.8 GB, about 1.5 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is Q2_K at 0.4 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35
Parameters752M
Layers24
Trained context262,144 tokens
KV cache (fp16)48 KB per token

Source

Quantized bymradermacher
FamilyIrwan
Licenseapache-2.0
Downloads (HF)870
Last updated2026-10-17

Frequently asked questions about Irwan LightOnOCR 3 0.8B Abliterated

How much VRAM does Irwan LightOnOCR 3 0.8B Abliterated need?

About 1.4 GB at Q4_K_M with an 8K context window. That is the 0.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Irwan LightOnOCR 3 0.8B Abliterated on an 8 GB GPU?

Yes, with F16: the 1.5 GB file needs about 2.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Irwan LightOnOCR 3 0.8B Abliterated GGUF?

It depends on the quant: Q2_K is 0.4 GB, Q3_K_S is 0.4 GB, Q3_K_M is 0.5 GB, Q3_K_L is 0.5 GB, Q4_K_S is 0.5 GB, IQ4_XS is 0.5 GB and 6 more.

What context length does Irwan LightOnOCR 3 0.8B Abliterated support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.6 GB at 32K.

Where can I download Irwan LightOnOCR 3 0.8B Abliterated GGUF?

The files are on Hugging Face in the mradermacher/Irwan-LightOnOCR-3-0.8B-abliterated-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Irwan LightOnOCR 3 0.8B Abliterated use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run this model on 8 GB VRAM?

Yes, use Q4_K_S or Q4_K_M which estimate 1.4 GB memory at 8 k context.

Does the model support vision tasks beyond OCR?

No, it is specifically designed for OCR and vision‑language reasoning.

What license applies to this model?

The model is released under Apache‑2.0.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

Other tiny models

All tiny models →