---

license: apache-2.0

base_model:

- cyberneurova/cyberneurova-Qwen3.8-27B

pipeline_tag: text-generation

tags:

- cyberneurova

- gguf

- llama.cpp

---

cyberneurova-Qwen3.8-27B — GGUF

Quantized GGUF builds of

cyberneurova/cyberneurova-Qwen3.8-27B

for running locally with llama.cpp, Ollama, LM Studio, Jan, or any

GGUF runtime. Uncensored, neutral, direct — tone is set entirely by your system prompt.

Capability is preserved: the base of these quants scores 11/12 (92%) pass@1 on

an executable coding benchmark (generated code run against hidden test cases).

💬 cyberneurova.ai  ·  ✉️ support@cyberneurova.com

Which file do I download?

Pick one file. Bigger = higher quality, more RAM/VRAM.

| File | Size | Notes |

|------|------|-------|

| cyberneurova-Qwen3.8-27B-Q2_K.gguf | ~10 GB | smallest, runs on modest hardware |

| cyberneurova-Qwen3.8-27B-Q3_K_M.gguf | ~13 GB | small |

| cyberneurova-Qwen3.8-27B-Q4_K_M.gguf | ~17 GB | recommended — best size/quality |

| cyberneurova-Qwen3.8-27B-Q5_K_M.gguf | ~19 GB | higher quality |

| cyberneurova-Qwen3.8-27B-Q6_K.gguf | ~22 GB | near-lossless |

| cyberneurova-Qwen3.8-27B-Q8_0.gguf | ~29 GB | highest-quality quant |

Run it (text)

llama.cpp

llama-cli -hf cyberneurova/cyberneurova-Qwen3.8-27B-GGUF:Q4_K_M -p "Hello"
# or a downloaded file:
llama-server -m cyberneurova-Qwen3.8-27B-Q4_K_M.gguf -c 8192

Ollama

ollama run hf.co/cyberneurova/cyberneurova-Qwen3.8-27B-GGUF:Q4_K_M

Run it with images (vision) 🖼️

This model is multimodal. To use images, add the vision projector

mmproj-F16.gguf (~0.9 GB) alongside any text quant:

llama-server \
  -m cyberneurova-Qwen3.8-27B-Q4_K_M.gguf \
  --mmproj mmproj-F16.gguf \
  -c 8192

Then send an image in the chat (llama.cpp's web UI, llama-mtmd-cli, or the

OpenAI-style API with an image_url). One mmproj-F16.gguf works with every

quant in this repo.

Notes

• Reasoning model: output includes a thinking trace, then the answer. Allow

enough tokens (1500+; 2000–4000 for code) or long answers get cut off.

• Send your own system prompt to fully control tone; leave it empty for the

built-in neutral default.

Disclaimer

Reduced built-in refusals. You are responsible for your use and for complying

with applicable laws. Provided as-is, without warranty.