dolphin-2.9.2-mixtral-8x22b needs about 88.0 GB of memory at Q4_K_M (a 85.6 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 12 quants, from 32.7 GB (IQ1_M) to 149.4 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 7.5 GB.
dolphin-2.9.2-mixtral-8x22b is a 141B-parameter Mistral model in GGUF format, quantized by bartowski. Its trained context length is 64K tokens. The repository has 12 quant files, from 32.7 GB (IQ1_M) to 149.4 GB (Q8_0).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_M | 32.7 GB | 35.1 GB | Too big | Get file → | |
| IQ2_XXS | 37.9 GB | 40.3 GB | Too big | Get file → | |
| IQ2_M | 46.7 GB | 49.1 GB | Too big | Get file → | |
| Q2_K | 52.1 GB (2 parts) | 54.5 GB | Too big | Get file → | |
| IQ3_XXS | 54.9 GB (2 parts) | 57.3 GB | Too big | Get file → | |
| Q3_K_S | 61.5 GB (2 parts) | 63.9 GB | Too big | Get file → | |
| IQ3_M | 64.5 GB (2 parts) | 66.9 GB | Too big | Get file → | |
| Q3_K_M | 67.8 GB (2 parts) | 70.2 GB | Too big | Get file → | |
| IQ4_XS | 75.5 GB (2 parts) | 77.9 GB | Too big | Get file → | |
| Q4_K_M | 85.6 GB (2 parts) | 88.0 GB | Too big | Get file → | |
| Q5_K_M | 100.0 GB (3 parts) | 102.4 GB | Too big | Get file → | |
| Q8_0 | 149.4 GB (4 parts) | 151.8 GB | Too big | Get file → |
What GPU do you need for dolphin-2.9.2-mixtral-8x22b?
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: the best quant that fits is IQ2_XXS (37.9 GB file, about 40.3 GB in use).
Which dolphin-2.9.2-mixtral-8x22b quant should you choose?
Q4_K_M is the usual starting point for dolphin-2.9.2-mixtral-8x22b: it balances file size and quality well. Here it is a 85.6 GB file.
The Q8_0 file is 149.4 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 IQ1_M at 32.7 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 dolphin-2.9.2-mixtral-8x22b
How much VRAM does dolphin-2.9.2-mixtral-8x22b need?
About 88.0 GB at Q4_K_M with an 8K context window. That is the 85.6 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 dolphin-2.9.2-mixtral-8x22b on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 35.1 GB, so you would need a larger GPU or CPU offloading.
What is the file size of dolphin-2.9.2-mixtral-8x22b GGUF?
It depends on the quant: IQ1_M is 32.7 GB, IQ2_XXS is 37.9 GB, IQ2_M is 46.7 GB, Q2_K is 52.1 GB, IQ3_XXS is 54.9 GB, Q3_K_S is 61.5 GB and 6 more.
What context length does dolphin-2.9.2-mixtral-8x22b support?
The GGUF header lists a trained context length of 65,536 tokens (64K). The KV cache grows with the context you use: about 1.9 GB at 8K and 7.5 GB at 32K.
Where can I download dolphin-2.9.2-mixtral-8x22b GGUF?
The files are on Hugging Face in the bartowski/dolphin-2.9.2-mixtral-8x22b-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 dolphin-2.9.2-mixtral-8x22b 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.