humanizer-12b needs about 21.0 GB of memory at Q4_K_M (a 7.6 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 6 quants, from 3.9 GB (IQ2_XS) to 23.8 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 51.5 GB.
Humanizer‑12B: a 12B multimodal model that rewrites AI drafts to read humanly
Humanizer‑12B is a 12B Gemma‑4 model that rewrites AI‑written drafts in English and Chinese. It preserves all factual details and adds no extra content. The model supports text, image, audio, and video inputs and runs locally via llama.cpp or the provided app.
Good at
- Rewriting AI drafts to read humanly
- Preserving factual details
- Multimodal input handling
Limits
- No vision support in the provided quant files
- Large memory requirement for higher precision quants
- Requires Apache‑2.0 license compliance
Best for
- Local draft rewriting
- Offline content editing
- Multimodal text generation
Who it’s for. Developers and content creators who need a local tool to transform AI‑generated text into human‑like prose while preserving facts. The model is suitable for offline use on laptops and mobile devices with sufficient VRAM.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XS | 3.9 GB | 17.3 GB | Too big | Get file → | |
| — | 5.6 GB | 19.0 GB | Too big | Get file → | |
| Q4_K_M | 7.6 GB | 21.0 GB | Too big | Get file → | |
| Q6_K | 10.0 GB | 23.4 GB | Too big | Get file → | |
| Q8_0 | 12.7 GB | 26.1 GB | Too big | Get file → | |
| BF16 | 23.8 GB | 37.2 GB | Too big | Get file → |
For 8 GB VRAM, use the IQ2_XS quant (3.89 GB file, ~17.3 GB memory at 8k context). For 12 GB VRAM, the Q4_K_M quant (7.63 GB file, ~21 GB memory) is a good balance. For 16 GB VRAM or more, the Q6_K or Q8_0 quants (10.03 GB or 12.67 GB files) provide higher precision.
What GPU do you need for humanizer-12b?
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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is Q6_K (10.0 GB file, about 23.4 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (12.7 GB file, about 26.1 GB in use).
- 48 GB VRAM: the best quant that fits is BF16 (23.8 GB file, about 37.2 GB in use).
Which humanizer-12b quant should you choose?
Q4_K_M is the usual starting point for humanizer-12b: it balances file size and quality well. Here it is a 7.6 GB file.
The Q8_0 file is 12.7 GB, about 1.7 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 IQ2_XS at 3.9 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
Source
Frequently asked questions about humanizer-12b
How much VRAM does humanizer-12b need?
About 21.0 GB at Q4_K_M with an 8K context window. That is the 7.6 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 humanizer-12b on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 17.3 GB, so you would need a larger GPU or CPU offloading.
What is the file size of humanizer-12b GGUF?
It depends on the quant: IQ2_XS is 3.9 GB, file is 5.6 GB, Q4_K_M is 7.6 GB, Q6_K is 10.0 GB, Q8_0 is 12.7 GB, BF16 is 23.8 GB.
What context length does humanizer-12b support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 12.9 GB at 8K and 51.5 GB at 32K.
Where can I download humanizer-12b GGUF?
The files are on Hugging Face in the ericthewiz/humanizer repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does humanizer-12b 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 Humanizer‑12B on 8 GB VRAM?
Yes, the IQ2_XS quant uses 3.89 GB and requires about 17.3 GB memory at 8k context, which fits within 8 GB VRAM when using efficient batching.
Does the model support image input?
Yes, Humanizer‑12B inherits multimodal capabilities from Gemma‑4, handling text and image inputs natively.
What license applies to this model?
The model is released under the Apache‑2.0 license, allowing free use and modification.