krea2_raw needs about 8.6 GB of memory at Q4_K_M (a 7.2 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 4 quants, from 7.2 GB (Q4_K_M) to 25.6 GB (BF16).
12.8B Krea-2 Raw model in GGUF format
Krea-2 Raw is a 12.8B parameter language model. It is available in multiple quantizations for use with llama.cpp and ComfyUI. The model is provided in FP8 weights and is compatible with lora training.
Good at
- Large context handling
- FP8 compatibility
- ComfyUI integration
Limits
- No vision support
- License unknown
- No mixture of experts
Best for
- Fine‑tuning with lora
- ComfyUI workflows
- FP8 inference
Who it’s for. Developers and researchers needing a 12.8B language model that can be quantized for efficient inference. The model is suitable for those who work with llama.cpp or ComfyUI and require FP8 weights.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 7.2 GB | 8.6 GB | Too big | Get file → | |
| Q6_K | 10.5 GB | 12.3 GB | Too big | Get file → | |
| Q8_0 | 13.6 GB | 15.8 GB | Too big | Get file → | |
| BF16 | 25.6 GB | 29.2 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_M (7.22 GB file, 8.6 GB memory). For 12 GB VRAM, Q6_K (10.52 GB file, 12.3 GB memory). For 16 GB VRAM, Q8_0 (13.63 GB file, 15.8 GB memory). BF16 requires 25.64 GB file and 29.2 GB memory.
What GPU do you need for krea2_raw?
These figures add the file size, an fp16 KV cache at an 8K context window 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: the best quant that fits is Q4_K_M (7.2 GB file, about 8.6 GB in use).
- 16 GB VRAM: the best quant that fits is Q8_0 (13.6 GB file, about 15.8 GB in use).
- 24 GB VRAM: the best quant that fits is Q8_0 (13.6 GB file, about 15.8 GB in use).
- 32 GB VRAM: the best quant that fits is BF16 (25.6 GB file, about 29.2 GB in use).
Which krea2_raw quant should you choose?
Q4_K_M is the usual starting point for krea2_raw: it balances file size and quality well. Here it is a 7.2 GB file.
The Q8_0 file is 13.6 GB, about 1.9 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.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about krea2_raw
How much VRAM does krea2_raw need?
About 8.6 GB at Q4_K_M with an 8K context window. That is the 7.2 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 krea2_raw on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 8.6 GB, so you would need a larger GPU or CPU offloading.
What is the file size of krea2_raw GGUF?
It depends on the quant: Q4_K_M is 7.2 GB, Q6_K is 10.5 GB, Q8_0 is 13.6 GB, BF16 is 25.6 GB.
Where can I download krea2_raw GGUF?
The files are on Hugging Face in the ChrisColeTech/krea-2-raw-FP8 repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does krea2_raw use?
Hugging Face lists the license as unknown. 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 Krea-2 Raw on 8 GB VRAM?
Yes, using the Q4_K_M quantization which requires 8.6 GB of VRAM at 8k context.
Is the model compatible with lora training?
The non-scaled FP8 weights are fully compatible with lora trainers.
Does the model support vision tasks?
No, the model does not have vision capabilities.