NuMarkdown-8B-Thinking needs about 5.7 GB of memory at Q4_K_M (a 4.7 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 25 quants, from 0.0 GB to 6.3 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.
NuMarkdown-8B-Thinking: a reasoning OCR VLM for document‑to‑Markdown conversion
NuMarkdown-8B-Thinking is a 7.62B parameter vision model fine‑tuned for reasoning OCR tasks. It generates thinking tokens to analyze document layout before producing clean Markdown. The model is trained on synthetic Doc→Reasoning→Markdown examples and a RL phase with a layout‑centric reward.
Good at
- Document layout reasoning
- Complex table extraction
- Markdown generation
Limits
- No language metadata
- Requires vision input
- Large context length 128000 tokens
Best for
- RAG document ingestion
- Automated PDF to Markdown pipelines
- Complex table conversion
Who it’s for. Developers building RAG systems that need accurate Markdown from PDFs. Researchers testing reasoning OCR models. Teams needing automated document conversion with layout awareness.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 1.0 GB | Fits | Get file → | |
| IQ1_S | 1.9 GB | 2.9 GB | Fits | Get file → | |
| IQ1_M | 2.0 GB | 3.0 GB | Fits | Get file → | |
| IQ2_XXS | 2.3 GB | 3.2 GB | Fits | Get file → | |
| IQ2_XS | 2.5 GB | 3.4 GB | Fits | Get file → | |
| IQ2_S | 2.6 GB | 3.6 GB | Fits | Get file → | |
| IQ2_M | 2.8 GB | 3.8 GB | Fits | Get file → | |
| Q2_K_S | 2.8 GB | 3.8 GB | Fits | Get file → | |
| Q2_K | 3.0 GB | 4.0 GB | Fits | Get file → | |
| IQ3_XXS | 3.1 GB | 4.1 GB | Fits | Get file → | |
| IQ3_XS | 3.3 GB | 4.3 GB | Fits | Get file → | |
| Q3_K_S | 3.5 GB | 4.5 GB | Fits | Get file → | |
| IQ3_S | 3.5 GB | 4.5 GB | Fits | Get file → | |
| IQ3_M | 3.6 GB | 4.5 GB | Fits | Get file → | |
| Q3_K_M | 3.8 GB | 4.8 GB | Fits | Get file → | |
| Q3_K_L | 4.1 GB | 5.1 GB | Fits | Get file → | |
| IQ4_XS | 4.2 GB | 5.2 GB | Fits | Get file → | |
| IQ4_NL | 4.4 GB | 5.4 GB | Fits | Get file → | |
| Q4_0 | 4.4 GB | 5.4 GB | Fits | Get file → | |
| Q4_K_S | 4.5 GB | 5.4 GB | Fits | Get file → | |
| Q4_K_M | 4.7 GB | 5.7 GB | Fits | Get file → | |
| Q4_1 | 4.9 GB | 5.8 GB | Fits | Get file → | |
| Q5_K_S | 5.3 GB | 6.3 GB | Fits | Get file → | |
| Q5_K_M | 5.4 GB | 6.4 GB | Fits | Get file → | |
| Q6_K | 6.3 GB | 7.2 GB | Tight | Get file → |
For 8GB VRAM, use IQ1_S (1.9 GB file, 2.9 GB memory at 8k context). For 12GB VRAM, IQ2_M (2.78 GB file, 3.8 GB memory). For 16GB VRAM, Q4_K_M (4.68 GB file, 5.7 GB memory). Estimates are approximate.
What GPU do you need for NuMarkdown-8B-Thinking?
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 Q2_K (3.0 GB file, about 4.0 GB in use).
- 6 GB VRAM: the best quant that fits is Q4_1 (4.9 GB file, about 5.8 GB in use).
- 8 GB VRAM: the best quant that fits is Q6_K (6.3 GB file, about 7.2 GB in use).
Which NuMarkdown-8B-Thinking quant should you choose?
Q4_K_M is the usual starting point for NuMarkdown-8B-Thinking: it balances file size and quality well. Here it is a 4.7 GB file.
If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about NuMarkdown-8B-Thinking
How much VRAM does NuMarkdown-8B-Thinking need?
About 5.7 GB at Q4_K_M with an 8K context window. That is the 4.7 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 NuMarkdown-8B-Thinking on an 8 GB GPU?
Yes, with Q6_K: the 6.3 GB file needs about 7.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of NuMarkdown-8B-Thinking GGUF?
It depends on the quant: file is 0.0 GB, IQ1_S is 1.9 GB, IQ1_M is 2.0 GB, IQ2_XXS is 2.3 GB, IQ2_XS is 2.5 GB, IQ2_S is 2.6 GB and 19 more.
What context length does NuMarkdown-8B-Thinking support?
The GGUF header lists a trained context length of 128,000 tokens (125K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.
Where can I download NuMarkdown-8B-Thinking GGUF?
The files are on Hugging Face in the mradermacher/NuMarkdown-8B-Thinking-fork-i1-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 NuMarkdown-8B-Thinking use?
Hugging Face lists the license as mit. 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 NuMarkdown-8B-Thinking on 8 GB VRAM?
Use the IQ1_S quant; it requires about 2.9 GB memory at 8k context, which fits within 8 GB VRAM.
Does this model support vision input?
Yes, it is a vision model and requires image input for OCR tasks.
Which quant is best for a 12 GB VRAM setup?
IQ2_M offers a good balance with 2.78 GB file size and 3.8 GB memory at 8k context.