QwenGram-0.8B REAL-15M R=1 needs about 1.5 GB of memory at Q4_K_M (a 0.6 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 4 quants, from 0.6 GB (Q4_K_M) to 1.6 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.7 GB.
QwenGram-0.8B: a compact Qwen3.5 derivative with a 15M reader for efficient inference
QwenGram-0.8B is a 0.8B parameter causal language model based on Qwen3.5. It includes a frozen Qwen3.5 backbone and a 15M reader for dynamic arbitration. The model is released under Apache‑2.0 and is designed for research and prototyping.
Good at
- Efficient inference with small VRAM
- Fast dynamic reader arbitration
- Research‑grade performance on text tasks
Limits
- Requires external PLE sidecar (~32 GB)
- No vision support in the GGUF file
- Large context tokens (262 144) may need more memory
Best for
- Fine‑tuning small language models
- Rapid prototyping of text generation
- Research on reader‑based arbitration
Who it’s for. Developers and researchers who need a lightweight Qwen3.5 variant for text‑only tasks and can manage the external PLE dependency. The model is suitable for experimentation and small‑scale deployment.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 0.6 GB | 1.5 GB | Fits | Get file → | |
| Q6_K | 0.7 GB | 1.6 GB | Fits | Get file → | |
| Q8_0 | 0.9 GB | 1.8 GB | Fits | Get file → | |
| BF16 | 1.6 GB | 2.5 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K_M (0.58 GB, ~1.5 GB VRAM). For 12 GB VRAM, Q6_K (0.69 GB, ~1.6 GB VRAM) is a good balance. If more memory is available, Q8_0 or BF16 can be used for higher precision.
What GPU do you need for QwenGram-0.8B REAL-15M R=1?
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 BF16 (1.6 GB file, about 2.5 GB in use).
Which QwenGram-0.8B REAL-15M R=1 quant should you choose?
Q4_K_M is the usual starting point for QwenGram-0.8B REAL-15M R=1: it balances file size and quality well. Here it is a 0.6 GB file.
The Q8_0 file is 0.9 GB, about 1.5 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.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about QwenGram-0.8B REAL-15M R=1
How much VRAM does QwenGram-0.8B REAL-15M R=1 need?
About 1.5 GB at Q4_K_M with an 8K context window. That is the 0.6 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 QwenGram-0.8B REAL-15M R=1 on an 8 GB GPU?
Yes, with BF16: the 1.6 GB file needs about 2.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of QwenGram-0.8B REAL-15M R=1 GGUF?
It depends on the quant: Q4_K_M is 0.6 GB, Q6_K is 0.7 GB, Q8_0 is 0.9 GB, BF16 is 1.6 GB.
What context length does QwenGram-0.8B REAL-15M R=1 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.7 GB at 32K.
Where can I download QwenGram-0.8B REAL-15M R=1 GGUF?
The files are on Hugging Face in the Ninnix96/Qwengram-0.8B repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does QwenGram-0.8B REAL-15M R=1 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 QwenGram-0.8B on 8 GB VRAM?
Yes, using the Q4_K_M quantization which requires about 1.5 GB VRAM at 8 k context.
Is vision support included?
No, the GGUF files contain only the text backbone; vision requires the external PLE sidecar.
What license applies to this model?
The model is released under the Apache‑2.0 license.