XwinXtended_v1.0_20B needs about 22.9 GB of memory at Q4_K_M (a 12.0 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 24 quants, from 0.0 GB to 16.4 GB (Q6_K).
XwinXtended_v1.0_20B is a 20B-parameter Xwinxtended model in GGUF format, quantized by mradermacher. Its trained context length is 4K tokens. The repository has 24 quant files, from 0.0 GB to 16.4 GB (Q6_K).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 5.7 GB | Fits | Get file → | |
| IQ1_S | 4.4 GB | 10.1 GB | Too big | Get file → | |
| IQ1_M | 4.8 GB | 10.5 GB | Too big | Get file → | |
| IQ2_XXS | 5.4 GB | 11.1 GB | Too big | Get file → | |
| IQ2_XS | 5.9 GB | 11.6 GB | Too big | Get file → | |
| IQ2_S | 6.4 GB | 12.1 GB | Too big | Get file → | |
| Q2_K_S | 6.8 GB | 12.5 GB | Too big | Get file → | |
| IQ2_M | 6.9 GB | 12.6 GB | Too big | Get file → | |
| Q2_K | 7.4 GB | 13.1 GB | Too big | Get file → | |
| IQ3_XXS | 7.6 GB | 13.3 GB | Too big | Get file → | |
| IQ3_XS | 8.2 GB | 13.9 GB | Too big | Get file → | |
| IQ3_S | 8.7 GB | 14.4 GB | Too big | Get file → | |
| Q3_K_S | 8.7 GB | 14.4 GB | Too big | Get file → | |
| IQ3_M | 9.2 GB | 14.9 GB | Too big | Get file → | |
| Q3_K_M | 9.7 GB | 15.4 GB | Too big | Get file → | |
| Q3_K_L | 10.6 GB | 16.3 GB | Too big | Get file → | |
| IQ4_XS | 10.7 GB | 16.4 GB | Too big | Get file → | |
| Q4_0 | 11.3 GB | 17.0 GB | Too big | Get file → | |
| Q4_K_S | 11.4 GB | 17.1 GB | Too big | Get file → | |
| Q4_K_M | 12.0 GB | 17.7 GB | Too big | Get file → | |
| Q4_1 | 12.5 GB | 18.2 GB | Too big | Get file → | |
| Q5_K_S | 13.8 GB | 19.5 GB | Too big | Get file → | |
| Q5_K_M | 14.2 GB | 19.9 GB | Too big | Get file → | |
| Q6_K | 16.4 GB | 22.1 GB | Too big | Get file → |
What GPU do you need for XwinXtended_v1.0_20B?
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: the best quant that fits is the main file (0.0 GB file, about 10.9 GB in use).
- 16 GB VRAM: the best quant that fits is IQ1_M (4.8 GB file, about 15.7 GB in use).
- 24 GB VRAM: the best quant that fits is Q4_1 (12.5 GB file, about 23.4 GB in use).
- 32 GB VRAM: the best quant that fits is Q6_K (16.4 GB file, about 27.3 GB in use).
Which XwinXtended_v1.0_20B quant should you choose?
Q4_K_M is the usual starting point for XwinXtended_v1.0_20B: it balances file size and quality well. Here it is a 12.0 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 XwinXtended_v1.0_20B
How much VRAM does XwinXtended_v1.0_20B need?
About 22.9 GB at Q4_K_M with an 8K context window. That is the 12.0 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 XwinXtended_v1.0_20B on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 10.9 GB, so you would need a larger GPU or CPU offloading.
What is the file size of XwinXtended_v1.0_20B GGUF?
It depends on the quant: file is 0.0 GB, IQ1_S is 4.4 GB, IQ1_M is 4.8 GB, IQ2_XXS is 5.4 GB, IQ2_XS is 5.9 GB, IQ2_S is 6.4 GB and 18 more.
What context length does XwinXtended_v1.0_20B support?
The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 10.4 GB at 8K and 5.2 GB at 4K.
Where can I download XwinXtended_v1.0_20B GGUF?
The files are on Hugging Face in the mradermacher/XwinXtended_v1.0_20B-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.