Sparse Llama 3.1 8B 2of4 needs about 6.5 GB of memory at Q4_K_M (a 4.9 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 24 quants, from 2.9 GB (IQ2_M) to 16.1 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.
Sparse Llama 3.1 8B 2of4 – 2:4 sparsity for efficient inference
Sparse Llama 3.1 8B 2of4 is a 8B Llama-3.1 model with 2:4 semi‑structured sparsity. It retains 50 % of weights and reduces compute by roughly 2×. The model is trained with knowledge distillation to recover accuracy loss.
Good at
- Efficient inference on CPUs
- Reduced model size
- Maintains dense model accuracy
Limits
- No vision support
- Requires 8 k context for memory estimates
- License llama3.1
Best for
- Edge deployments
- Cost‑effective scaling
- CPU‑only inference
Who it’s for. Engineers and researchers looking to deploy Llama‑3.1 models on resource‑constrained hardware. The 2:4 sparsity allows significant savings while keeping performance close to the dense baseline.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_M | 2.9 GB | 4.5 GB | Fits | Get file → | |
| Q2_K | 3.2 GB | 4.8 GB | Fits | Get file → | |
| IQ3_XS | 3.5 GB | 5.1 GB | Fits | Get file → | |
| Q3_K_S | 3.7 GB | 5.2 GB | Fits | Get file → | |
| Q2_K_L | 3.7 GB | 5.3 GB | Fits | Get file → | |
| IQ3_M | 3.8 GB | 5.4 GB | Fits | Get file → | |
| Q3_K_M | 4.0 GB | 5.6 GB | Fits | Get file → | |
| Q3_K_L | 4.3 GB | 5.9 GB | Fits | Get file → | |
| IQ4_XS | 4.4 GB | 6.0 GB | Fits | Get file → | |
| Q4_0 | 4.7 GB | 6.2 GB | Fits | Get file → | |
| Q4_0 | 4.7 GB | 6.2 GB | Fits | Get file → | |
| Q4_0 | 4.7 GB | 6.2 GB | Fits | Get file → | |
| Q4_0 | 4.7 GB | 6.2 GB | Fits | Get file → | |
| Q4_K_S | 4.7 GB | 6.3 GB | Fits | Get file → | |
| Q3_K_XL | 4.8 GB | 6.4 GB | Fits | Get file → | |
| Q4_K_M | 4.9 GB | 6.5 GB | Fits | Get file → | |
| Q4_K_L | 5.3 GB | 6.9 GB | Fits | Get file → | |
| Q5_K_S | 5.6 GB | 7.2 GB | Fits | Get file → | |
| Q5_K_M | 5.7 GB | 7.3 GB | Tight | Get file → | |
| Q5_K_L | 6.1 GB | 7.6 GB | Tight | Get file → | |
| Q6_K | 6.6 GB | 8.2 GB | Too big | Get file → | |
| Q6_K_L | 6.9 GB | 8.4 GB | Too big | Get file → | |
| Q8_0 | 8.5 GB | 10.1 GB | Too big | Get file → | |
| F16 | 16.1 GB | 17.6 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_0 (file size 4.66 GB, memory 6.2 GB). For 12 GB VRAM, Q5_K_M (file size 5.73 GB, memory 7.3 GB) or Q6_K (file size 6.6 GB, memory 8.2 GB). For 16 GB VRAM, Q6_K_L (file size 6.85 GB, memory 8.4 GB) or Q8_0 (file size 8.54 GB, memory 10.1 GB).
What GPU do you need for Sparse Llama 3.1 8B 2of4?
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: the best quant that fits is IQ4_XS (4.4 GB file, about 6.0 GB in use).
- 8 GB VRAM: the best quant that fits is Q5_K_L (6.1 GB file, about 7.6 GB in use).
- 12 GB VRAM: the best quant that fits is Q8_0 (8.5 GB file, about 10.1 GB in use).
- 16 GB VRAM: the best quant that fits is Q8_0 (8.5 GB file, about 10.1 GB in use).
- 24 GB VRAM: the best quant that fits is F16 (16.1 GB file, about 17.6 GB in use).
Which Sparse Llama 3.1 8B 2of4 quant should you choose?
Q4_K_M is the usual starting point for Sparse Llama 3.1 8B 2of4: it balances file size and quality well. Here it is a 4.9 GB file.
The Q8_0 file is 8.5 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_M at 2.9 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 Sparse Llama 3.1 8B 2of4
How much VRAM does Sparse Llama 3.1 8B 2of4 need?
About 6.5 GB at Q4_K_M with an 8K context window. That is the 4.9 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 Sparse Llama 3.1 8B 2of4 on an 8 GB GPU?
Yes, with Q5_K_L: the 6.1 GB file needs about 7.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Sparse Llama 3.1 8B 2of4 GGUF?
It depends on the quant: IQ2_M is 2.9 GB, Q2_K is 3.2 GB, IQ3_XS is 3.5 GB, Q3_K_S is 3.7 GB, Q2_K_L is 3.7 GB, IQ3_M is 3.8 GB and 18 more.
What context length does Sparse Llama 3.1 8B 2of4 support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.
Where can I download Sparse Llama 3.1 8B 2of4 GGUF?
The files are on Hugging Face in the bartowski/Sparse-Llama-3.1-8B-2of4-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 Sparse Llama 3.1 8B 2of4 use?
Hugging Face lists the license as llama3.1. 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, the Q4_0 quant uses 4.66 GB file size and requires about 6.2 GB of VRAM at 8 k context.
Does this model support vision tasks?
No, it only accepts text input and produces text output.
What license applies to this model?
The model is licensed under llama3.1.