A.X 4.0 Imatrix needs about 50.0 GB of memory at Q4_K_M (a 46.9 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 8 quants, from 22.8 GB (TQ1_0) to 53.8 GB (Q5_K_M). Longer context needs more: at 32K tokens the KV cache alone adds about 10.7 GB.
A.X 4.0 Imatrix is a 71.9B-parameter model in GGUF format, quantized by cookieshake. Its trained context length is 128K tokens. The repository has 8 quant files, from 22.8 GB (TQ1_0) to 53.8 GB (Q5_K_M).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 22.8 GB | 26.0 GB | Too big | Get file → | |
| IQ1_M | 23.3 GB | 26.5 GB | Too big | Get file → | |
| IQ2_XXS | 25.1 GB | 28.3 GB | Too big | Get file → | |
| IQ2_M | 28.9 GB | 32.1 GB | Too big | Get file → | |
| Q2_K_S | 29.1 GB | 32.3 GB | Too big | Get file → | |
| Q2_K | 29.3 GB | 32.5 GB | Too big | Get file → | |
| Q4_K_M | 46.9 GB | 50.0 GB | Too big | Get file → | |
| Q5_K_M | 53.8 GB (2 parts) | 57.0 GB | Too big | Get file → |
What GPU do you need for A.X 4.0 Imatrix?
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: the best quant that fits is IQ2_XXS (25.1 GB file, about 28.3 GB in use).
- 48 GB VRAM: the best quant that fits is Q2_K (29.3 GB file, about 32.5 GB in use).
Which A.X 4.0 Imatrix quant should you choose?
Q4_K_M is the usual starting point for A.X 4.0 Imatrix: it balances file size and quality well. Here it is a 46.9 GB file.
If memory is tight, the smallest file is TQ1_0 at 22.8 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about A.X 4.0 Imatrix
How much VRAM does A.X 4.0 Imatrix need?
About 50.0 GB at Q4_K_M with an 8K context window. That is the 46.9 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 A.X 4.0 Imatrix on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 26.0 GB, so you would need a larger GPU or CPU offloading.
What is the file size of A.X 4.0 Imatrix GGUF?
It depends on the quant: TQ1_0 is 22.8 GB, IQ1_M is 23.3 GB, IQ2_XXS is 25.1 GB, IQ2_M is 28.9 GB, Q2_K_S is 29.1 GB, Q2_K is 29.3 GB and 2 more.
What context length does A.X 4.0 Imatrix support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 2.7 GB at 8K and 10.7 GB at 32K.
Where can I download A.X 4.0 Imatrix GGUF?
The files are on Hugging Face in the cookieshake/A.X-4.0-Imatrix-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 A.X 4.0 Imatrix 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.