---

license: apache-2.0

base_model: davidnichols-ops/claude-yolo-vibes-v5

tags:

- gguf

- qwen2.5-coder

- ollama

- llama.cpp

- text-generation

pipeline_tag: text-generation

---

claude-yolo-vibes-v5 — GGUF

GGUF quantizations of davidnichols-ops/claude-yolo-vibes-v5 (Qwen2.5-Coder-7B + distillation SFT + personality polish + DPO) for llama.cpp / Ollama / LM Studio.

| File | Quant | Size | Quality | Use |

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

| claude-yolo-vibes-v5-Q4_K_M.gguf | Q4_K_M | 4.7 GB | Good | Ollama default — recommended |

| claude-yolo-vibes-v5-Q5_K_M.gguf | Q5_K_M | 5.4 GB | Better | When you have the RAM |

| claude-yolo-vibes-v5-Q8_0.gguf | Q8_0 | 8.1 GB | Near-lossless | Max quality, 8+ GB RAM |

Ollama

ollama pull davidnicholsops/claude-yolo-vibes-v5

Or import directly from the GGUF:

# download the Q4_K_M file, then:
ollama create claude-yolo-vibes-v5 -f Modelfile

# Modelfile
FROM ./claude-yolo-vibes-v5-Q4_K_M.gguf
TEMPLATE """{{- if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{- end }}<|im_start|>user
{{ .Prompt }}<|im_end|>
<|im_start|>assistant
"""
SYSTEM """You are claude-yolo-vibes, a coding assistant with a sharp, playful edge. You write working code first and make jokes second."""
PARAMETER temperature 0.7
PARAMETER stop "<|im_end|>"

llama.cpp

llama-cli -m claude-yolo-vibes-v5-Q4_K_M.gguf \
  -p "Write a Python function that computes fibonacci with memoization." -n 256

Verification

• All quants converted from the same davidnichols-ops/claude-yolo-vibes-v5 safetensors.
• Q8_0 smoke-tested with llama.cpp: loads, chat template intact, generates correct code.
• HumanEval (full-precision model): 85.4% (140/164) — +1.0 vs v4-dpo measured on the same harness.

License

Apache 2.0.