Moe X33 needs about 38.0 GB of memory at Q4_K_M (a 35.4 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 20 quants, from 21.6 GB (Q2_K) to 44.2 GB (Q5_1).
Moe X33 is a 58.9B-parameter Moe model in GGUF format, quantized by RichardErkhov. Its trained context length is 16K tokens. The repository has 20 quant files, from 21.6 GB (Q2_K) to 44.2 GB (Q5_1).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 21.6 GB | 24.2 GB | Too big | Get file → | |
| IQ3_XS | 24.1 GB | 26.7 GB | Too big | Get file → | |
| Q3_K_S | 25.4 GB | 28.0 GB | Too big | Get file → | |
| IQ3_S | 25.5 GB | 28.1 GB | Too big | Get file → | |
| IQ3_M | 26.2 GB | 28.7 GB | Too big | Get file → | |
| Q3_K | 28.3 GB | 30.9 GB | Too big | Get file → | |
| Q3_K_M | 28.3 GB | 30.9 GB | Too big | Get file → | |
| Q3_K_L | 30.8 GB | 33.3 GB | Too big | Get file → | |
| Q4_K_S | 31.4 GB | 34.0 GB | Too big | Get file → | |
| IQ4_XS | 31.8 GB | 34.3 GB | Too big | Get file → | |
| Q4_0 | 33.2 GB | 35.8 GB | Too big | Get file → | |
| IQ4_NL | 33.5 GB | 36.1 GB | Too big | Get file → | |
| Q4_K | 35.4 GB | 38.0 GB | Too big | Get file → | |
| Q4_K_M | 35.4 GB | 38.0 GB | Too big | Get file → | |
| Q4_1 | 36.9 GB | 39.5 GB | Too big | Get file → | |
| Q5_0 | 40.6 GB (2 parts) | 43.1 GB | Too big | Get file → | |
| Q5_K_S | 40.6 GB (2 parts) | 43.1 GB | Too big | Get file → | |
| Q5_K | 41.7 GB (2 parts) | 44.3 GB | Too big | Get file → | |
| Q5_K_M | 41.7 GB (2 parts) | 44.3 GB | Too big | Get file → | |
| Q5_1 | 44.2 GB (2 parts) | 46.8 GB | Too big | Get file → |
What GPU do you need for Moe X33?
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 Q3_K (28.3 GB file, about 30.9 GB in use).
- 48 GB VRAM: the best quant that fits is Q5_1 (44.2 GB file, about 46.8 GB in use).
Which Moe X33 quant should you choose?
Q4_K_M is the usual starting point for Moe X33: it balances file size and quality well. Here it is a 35.4 GB file.
If memory is tight, the smallest file is Q2_K at 21.6 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 Moe X33
How much VRAM does Moe X33 need?
About 38.0 GB at Q4_K_M with an 8K context window. That is the 35.4 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 Moe X33 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 24.2 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Moe X33 GGUF?
It depends on the quant: Q2_K is 21.6 GB, IQ3_XS is 24.1 GB, Q3_K_S is 25.4 GB, IQ3_S is 25.5 GB, IQ3_M is 26.2 GB, Q3_K is 28.3 GB and 14 more.
What context length does Moe X33 support?
The GGUF header lists a trained context length of 16,384 tokens (16K). The KV cache grows with the context you use: about 2.1 GB at 8K and 4.2 GB at 16K.
Where can I download Moe X33 GGUF?
The files are on Hugging Face in the RichardErkhov/senseable_-_moe-x33-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 Moe X33 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.