Theprint Moe 8x3 0126 needs about 12.5 GB of memory at Q4_K_M (a 11.1 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 12 quants, from 6.9 GB (Q2_K) to 36.0 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 3.8 GB.
18B Mixture of Experts Llama model
Theprint Moe 8x3 0126 is an 18B parameter Mixture of Experts model built on Llama architecture. It combines 8 specialized 3B experts with 2 experts active per token by default. The model supports a 131072 token context window.
Good at
- Tool usage
- Large context handling
- Expertise switching
Limits
- No vision support
- Requires 8k context for memory estimates
- Apache-2.0 license
Best for
- Chatbots needing deep context
- Complex tool‑integrated tasks
- Fine‑tuned expert inference
Who it’s for. Developers and researchers who need a large‑scale Llama model with Mixture of Experts capability and a 131072 token context window. The model is suitable for applications that can benefit from expert specialization and tool integration.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 6.9 GB | 8.4 GB | Too big | Get file → | |
| Q3_K_S | 8.1 GB | 9.5 GB | Too big | Get file → | |
| Q3_K_M | 8.9 GB | 10.3 GB | Too big | Get file → | |
| Q3_K_L | 9.5 GB | 10.9 GB | Too big | Get file → | |
| IQ4_XS | 9.9 GB | 11.4 GB | Too big | Get file → | |
| Q4_0 | 10.3 GB | 11.8 GB | Too big | Get file → | |
| Q4_K_M | 11.1 GB | 12.5 GB | Too big | Get file → | |
| Q4_1 | 11.4 GB | 12.9 GB | Too big | Get file → | |
| Q5_K_M | 12.9 GB | 14.3 GB | Too big | Get file → | |
| Q6_K | 14.8 GB | 16.3 GB | Too big | Get file → | |
| Q8_0 | 19.1 GB | 20.6 GB | Too big | Get file → | |
| F16 | 36.0 GB | 37.5 GB | Too big | Get file → |
For 8 GB VRAM, use Q2_K (6.91 GB file, 8.4 GB memory). For 12 GB VRAM, Q4_K_M (11.09 GB file, 12.5 GB memory) is a good balance. For 20 GB VRAM, Q8_0 (19.15 GB file, 20.6 GB memory) provides full precision.
What GPU do you need for Theprint Moe 8x3 0126?
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 Q4_0 (10.3 GB file, about 11.8 GB in use).
- 16 GB VRAM: the best quant that fits is Q5_K_M (12.9 GB file, about 14.3 GB in use).
- 24 GB VRAM: the best quant that fits is Q8_0 (19.1 GB file, about 20.6 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (19.1 GB file, about 20.6 GB in use).
- 48 GB VRAM: the best quant that fits is F16 (36.0 GB file, about 37.5 GB in use).
Which Theprint Moe 8x3 0126 quant should you choose?
Q4_K_M is the usual starting point for Theprint Moe 8x3 0126: it balances file size and quality well. Here it is a 11.1 GB file.
The Q8_0 file is 19.1 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 Q2_K at 6.9 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 Theprint Moe 8x3 0126
How much VRAM does Theprint Moe 8x3 0126 need?
About 12.5 GB at Q4_K_M with an 8K context window. That is the 11.1 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 Theprint Moe 8x3 0126 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 8.4 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Theprint Moe 8x3 0126 GGUF?
It depends on the quant: Q2_K is 6.9 GB, Q3_K_S is 8.1 GB, Q3_K_M is 8.9 GB, Q3_K_L is 9.5 GB, IQ4_XS is 9.9 GB, Q4_0 is 10.3 GB and 6 more.
What context length does Theprint Moe 8x3 0126 support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.9 GB at 8K and 3.8 GB at 32K.
Where can I download Theprint Moe 8x3 0126 GGUF?
The files are on Hugging Face in the theprint/theprint-moe-8x3-0126-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 Theprint Moe 8x3 0126 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.
Can I run this model on 8 GB VRAM?
Yes, the Q2_K quant uses 6.91 GB file size and requires about 8.4 GB memory at 8k context.
Does the model support vision?
No, the model does not have vision capabilities.
How many experts are active per token by default?
Two experts are active per token by default, up to four can be configured.