Huihui3.5 67B A3B APEX needs about 44.8 GB of memory at its main file (a 43.6 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 8 quants, from 27.3 GB to 133.9 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 2.7 GB.
Huihui3.5 67B A3B APEX – a large MoE model with adaptive precision quantization
Huihui3.5 67B A3B APEX is a 67B‑parameter Mixture‑of‑Experts model based on qwen3_5_moe architecture. It supports 262144‑token context and uses APEX quantization to reduce memory usage while preserving expert routing. The model is released under Apache‑2.0.
Good at
- Large‑context reasoning
- Expert routing for diverse tasks
- Adaptive precision compression
Limits
- Requires >28 GB VRAM for 8 k context
- No vision support
- Active 3 B parameters per token
Best for
- Long‑form content generation
- Complex code synthesis
- Multilingual dialogue
Who it’s for. Developers and researchers needing high‑capacity MoE models for local inference. Suitable for environments with ample GPU memory and a need for fine‑grained expert control.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 27.3 GB | 28.4 GB | Too big | Get file → | |
| — | 33.0 GB | 34.2 GB | Too big | Get file → | |
| — | 33.0 GB | 34.2 GB | Too big | Get file → | |
| — | 43.6 GB | 44.8 GB | Too big | Get file → | |
| — | 43.6 GB | 44.8 GB | Too big | Get file → | |
| — | 49.3 GB | 50.5 GB | Too big | Get file → | |
| — | 49.3 GB | 50.5 GB | Too big | Get file → | |
| F16 | 133.9 GB | 135.1 GB | Too big | Get file → |
For 8 k context, the F16 quant uses ~135.1 GB VRAM. If memory is limited, choose one of the unknown quant files; the smallest requires ~28.4 GB VRAM and the largest ~50.5 GB. Estimates are approximate.
What GPU do you need for Huihui3.5 67B A3B APEX?
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 the main file (27.3 GB file, about 28.4 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (43.6 GB file, about 44.8 GB in use).
Which Huihui3.5 67B A3B APEX quant should you choose?
If memory is tight, the smallest file is the first one in the table at 27.3 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Huihui3.5 67B A3B APEX
How much VRAM does Huihui3.5 67B A3B APEX need?
About 44.8 GB at its main file with an 8K context window. That is the 43.6 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 Huihui3.5 67B A3B APEX on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 28.4 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Huihui3.5 67B A3B APEX GGUF?
It depends on the quant: file is 27.3 GB, file is 33.0 GB, file is 33.0 GB, file is 43.6 GB, file is 43.6 GB, file is 49.3 GB and 2 more.
What context length does Huihui3.5 67B A3B APEX support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.7 GB at 8K and 2.7 GB at 32K.
Where can I download Huihui3.5 67B A3B APEX GGUF?
The files are on Hugging Face in the mudler/Huihui3.5-67B-A3B-APEX-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 Huihui3.5 67B A3B APEX 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?
No. The smallest quant requires at least 28.4 GB VRAM for 8 k context. 8 GB is insufficient.
What is the maximum context length?
The model supports up to 262144 tokens of context.
Does it support vision?
No, the model has no vision capabilities.