PXA bench files needs about 15.9 GB of memory at its main file (a 14.7 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 6 quants, from 10.7 GB to 19.2 GB. Longer context needs more: at 32K tokens the KV cache alone adds about 2.7 GB.
PXA benchmark files – 34.7B Qwen‑35Moe model with vision, reasoning, and tools
PXA benchmark files are a set of 34.7B parameter Qwen‑35Moe models quantized for benchmarking. They support reasoning, tools, and vision tasks. The models are released under the Apache‑2.0 license and are available via the PXA engine.
Good at
- Reasoning
- Tools integration
- Vision tasks
Limits
- No stock llama.cpp support
- Large context size requires high VRAM
- No explicit language support listed
Best for
- Benchmarking cross‑engine performance
- Testing vision capabilities
- Evaluating reasoning and tool use
Who it’s for. Researchers and developers who need reproducible benchmark models for cross‑engine comparison. The files are intended for use with the PXA engine and are not compatible with standard llama.cpp setups.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 10.7 GB | 11.8 GB | Too big | Get file → | |
| — | 14.1 GB | 15.3 GB | Too big | Get file → | |
| — | 14.7 GB | 15.9 GB | Too big | Get file → | |
| — | 18.7 GB | 19.9 GB | Too big | Get file → | |
| MXFP4 | 19.1 GB | 20.3 GB | Too big | Get file → | |
| — | 19.2 GB | 20.3 GB | Too big | Get file → |
For 8 GB VRAM, the MXFP4 quant file (19.11 GB) is the smallest available and will fit with 20.3 GB estimated memory at 8 k context. If more VRAM is available, the 10.67 GB file (11.8 GB memory) offers a lighter option. Larger files exceed typical consumer GPU limits.
What GPU do you need for PXA bench files?
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 the main file (10.7 GB file, about 11.8 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (14.7 GB file, about 15.9 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (19.2 GB file, about 20.3 GB in use).
Which PXA bench files quant should you choose?
If memory is tight, the smallest file is the first one in the table at 10.7 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 PXA bench files
How much VRAM does PXA bench files need?
About 15.9 GB at its main file with an 8K context window. That is the 14.7 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 PXA bench files on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 11.8 GB, so you would need a larger GPU or CPU offloading.
What is the file size of PXA bench files GGUF?
It depends on the quant: file is 10.7 GB, file is 14.1 GB, file is 14.7 GB, file is 18.7 GB, MXFP4 is 19.1 GB, file is 19.2 GB.
What context length does PXA bench files 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 PXA bench files GGUF?
The files are on Hugging Face in the poisonxa/PXA-bench-files-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 PXA bench files 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 the PXA benchmark files on a GPU with 8 GB VRAM?
Only the MXFP4 quant file (19.11 GB) will fit, requiring about 20.3 GB memory at 8 k context. Smaller files still exceed 8 GB VRAM.
Do these models support vision tasks?
Yes, the detected abilities list includes vision. However, they require the PXA engine, not standard llama.cpp.
What license applies to these files?
They are released under the Apache‑2.0 license.