micro-qwen4exp-realqsa-mtp needs about 0.8 GB of memory at Q4_K_M (a 0.1 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 4 quants, from 0.1 GB (Q4_K_M) to 0.4 GB. Longer context needs more: at 32K tokens the KV cache alone adds about 0.9 GB.
Micro Qwen4exp Random Debug Model
A tiny, randomly initialized GGUF model for debugging Qwen4exp architecture. It contains vision path and MTP head but produces garbage text. Designed for profiling and testing llama.cpp.
Good at
- Speed profiling
- Graph build testing
- Memory management debugging
Limits
- Random weights – not for inference
- No real language support
- Vision path only for testing
Best for
- Profiling llama.cpp operations
- Testing graph layout
- CI pipeline coverage
Who it’s for. Engineers and developers working on Qwen4exp, MTP, and vision paths in llama.cpp. Useful for debugging, profiling, and CI testing.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 0.1 GB | 0.8 GB | Fits | Get file → | |
| Q4_K_M | 0.1 GB | 0.9 GB | Fits | Get file → | |
| — | 0.1 GB | 0.9 GB | Fits | Get file → | |
| — | 0.4 GB | 1.1 GB | Fits | Get file → |
For 8 k context, the 0.05 GB Q4_K_M file uses about 0.8 GB VRAM. If more memory is available, the 0.14 GB Q4_K_M file uses about 0.9 GB. Larger files use 0.9–1.1 GB VRAM.
What GPU do you need for micro-qwen4exp-realqsa-mtp?
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: the best quant that fits is the main file (0.4 GB file, about 1.1 GB in use).
Which micro-qwen4exp-realqsa-mtp quant should you choose?
Q4_K_M is the usual starting point for micro-qwen4exp-realqsa-mtp: it balances file size and quality well. Here it is a 0.1 GB file.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about micro-qwen4exp-realqsa-mtp
How much VRAM does micro-qwen4exp-realqsa-mtp need?
About 0.8 GB at Q4_K_M with an 8K context window. That is the 0.1 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 micro-qwen4exp-realqsa-mtp on an 8 GB GPU?
Yes, with the smaller files: the 0.4 GB file needs about 1.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of micro-qwen4exp-realqsa-mtp GGUF?
It depends on the quant: Q4_K_M is 0.1 GB, Q4_K_M is 0.1 GB, file is 0.1 GB, file is 0.4 GB.
What context length does micro-qwen4exp-realqsa-mtp support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.2 GB at 8K and 0.9 GB at 32K.
Where can I download micro-qwen4exp-realqsa-mtp GGUF?
The files are on Hugging Face in the Lynxpda/micro-qwen4exp repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does micro-qwen4exp-realqsa-mtp use?
Hugging Face lists the license as mit. 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 smallest Q4_K_M file uses about 0.8 GB VRAM at 8 k context, well within 8 GB.
Is this model suitable for inference?
No. All weights are random noise; the model is for debugging and profiling only.
What is the main purpose of this model?
To profile and test Qwen4exp architecture, MTP, and vision paths in llama.cpp.