Huihui MoE 23B A4B Abliterated needs about 16.0 GB of memory at Q4_K_M (a 14.2 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 16 quants, from 8.8 GB (Q2_K) to 19.1 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 4.8 GB.
Huihui MoE 23B A4B Abliterated is a 23.2B-parameter Huihui model in GGUF format, quantized by mradermacher. Its trained context length is 40K tokens. The repository has 16 quant files, from 8.8 GB (Q2_K) to 19.1 GB (Q6_K).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 8.8 GB | 10.5 GB | Too big | Get file → | |
| IQ3_XXS | 9.3 GB | 11.0 GB | Too big | Get file → | |
| IQ3_XS | 9.8 GB | 11.5 GB | Too big | Get file → | |
| IQ3_S | 10.4 GB | 12.1 GB | Too big | Get file → | |
| Q3_K_S | 10.4 GB | 12.1 GB | Too big | Get file → | |
| IQ3_M | 10.8 GB | 12.5 GB | Too big | Get file → | |
| Q3_K_M | 11.4 GB | 13.1 GB | Too big | Get file → | |
| Q3_K_L | 12.1 GB | 13.8 GB | Too big | Get file → | |
| IQ4_XS | 12.6 GB | 14.3 GB | Too big | Get file → | |
| Q4_0 | 13.3 GB | 15.0 GB | Too big | Get file → | |
| Q4_K_S | 13.4 GB | 15.1 GB | Too big | Get file → | |
| Q4_K_M | 14.2 GB | 16.0 GB | Too big | Get file → | |
| Q4_1 | 14.7 GB | 16.4 GB | Too big | Get file → | |
| Q5_K_S | 16.1 GB | 17.8 GB | Too big | Get file → | |
| Q5_K_M | 16.6 GB | 18.3 GB | Too big | Get file → | |
| Q6_K | 19.1 GB | 20.8 GB | Too big | Get file → |
What GPU do you need for Huihui MoE 23B A4B Abliterated?
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 IQ3_XS (9.8 GB file, about 11.5 GB in use).
- 16 GB VRAM: the best quant that fits is Q4_K_M (14.2 GB file, about 16.0 GB in use).
- 24 GB VRAM: the best quant that fits is Q6_K (19.1 GB file, about 20.8 GB in use).
Which Huihui MoE 23B A4B Abliterated quant should you choose?
Q4_K_M is the usual starting point for Huihui MoE 23B A4B Abliterated: it balances file size and quality well. Here it is a 14.2 GB file.
If memory is tight, the smallest file is Q2_K at 8.8 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 Huihui MoE 23B A4B Abliterated
How much VRAM does Huihui MoE 23B A4B Abliterated need?
About 16.0 GB at Q4_K_M with an 8K context window. That is the 14.2 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 Huihui MoE 23B A4B Abliterated on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 10.5 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Huihui MoE 23B A4B Abliterated GGUF?
It depends on the quant: Q2_K is 8.8 GB, IQ3_XXS is 9.3 GB, IQ3_XS is 9.8 GB, IQ3_S is 10.4 GB, Q3_K_S is 10.4 GB, IQ3_M is 10.8 GB and 10 more.
What context length does Huihui MoE 23B A4B Abliterated support?
The GGUF header lists a trained context length of 40,960 tokens (40K). The KV cache grows with the context you use: about 1.2 GB at 8K and 4.8 GB at 32K.
Where can I download Huihui MoE 23B A4B Abliterated GGUF?
The files are on Hugging Face in the mradermacher/Huihui-MoE-23B-A4B-abliterated-i1-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 Huihui MoE 23B A4B Abliterated 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.