shisa v2 llama3.1 405b needs about 247.8 GB of memory at Q4_K_M (a 243.1 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 6 quants, from 107.3 GB (IQ2_XXS) to 403.2 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 16.9 GB.
shisa v2 llama3.1 405b is a 406B-parameter Llama model in GGUF format, quantized by shisa-ai. Its trained context length is 128K tokens. The repository has 6 quant files, from 107.3 GB (IQ2_XXS) to 403.2 GB (Q8_0).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XXS | 107.3 GB (3 parts) | 112.0 GB | Too big | Get file → | |
| IQ3_XS | 166.1 GB (4 parts) | 170.8 GB | Too big | Get file → | |
| IQ3_M | 181.7 GB (5 parts) | 186.5 GB | Too big | Get file → | |
| IQ4_XS | 216.6 GB (5 parts) | 221.3 GB | Too big | Get file → | |
| Q4_K_M | 243.1 GB (6 parts) | 247.8 GB | Too big | Get file → | |
| Q8_0 | 403.2 GB (9 parts) | 407.9 GB | Too big | Get file → |
What GPU do you need for shisa v2 llama3.1 405b?
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: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which shisa v2 llama3.1 405b quant should you choose?
Q4_K_M is the usual starting point for shisa v2 llama3.1 405b: it balances file size and quality well. Here it is a 243.1 GB file.
The Q8_0 file is 403.2 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_XXS at 107.3 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 shisa v2 llama3.1 405b
How much VRAM does shisa v2 llama3.1 405b need?
About 247.8 GB at Q4_K_M with an 8K context window. That is the 243.1 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 shisa v2 llama3.1 405b on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 112.0 GB, so you would need a larger GPU or CPU offloading.
What is the file size of shisa v2 llama3.1 405b GGUF?
It depends on the quant: IQ2_XXS is 107.3 GB, IQ3_XS is 166.1 GB, IQ3_M is 181.7 GB, IQ4_XS is 216.6 GB, Q4_K_M is 243.1 GB, Q8_0 is 403.2 GB.
What context length does shisa v2 llama3.1 405b support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 4.2 GB at 8K and 16.9 GB at 32K.
Where can I download shisa v2 llama3.1 405b GGUF?
The files are on Hugging Face in the shisa-ai/shisa-v2-llama3.1-405b-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 shisa v2 llama3.1 405b 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.