---

license: apache-2.0

base_model:

• WineryLabs/Winery-Qwen3.5-9B-Atelier-GGUF
• WineryLabs/Winery-Qwen3.5-9B-Fable-Reserve-GGUF

base_model_relation: merge

library_name: gguf

pipeline_tag: text-generation

tags: [gguf, merge, slerp, qwen3.5, winery, llama.cpp, web-design, threejs, frontend, html]

---

<p align="center"><a href="https://huggingface.co/WineryLabs"><img src="https://winery-api-zandy.zocomputer.io/card/m/fatelier.svg" alt="Winery Qwen3.5-9B-Fable-Atelier" width="100%"/></a></p>

🍷 Winery Qwen3.5 9B · Fable Atelier

Two Winery vintages blended into one web design + 3D specialist. It takes the front-end instincts of Atelier and adds a measure of Fable Reserve for cleaner layout and type. Give it a brief and it writes a complete single-file site: landing pages, glass UI, scroll animations and live three.js scenes. 9.5 GB at Q8_0.

👉 Fable Atelier's own Space was designed and coded by this model.

Assemblage

| Parent | Share | What it brings |

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

| Winery 9B Atelier | 70% | region-weighted merge of six web/UI/coding tunes |

| Winery 9B Fable Reserve | 30% | bf16 soup of five Claude Fable 5 distills (80.1 quick eval) |

SLERP with t = 0.7 on every layer, embeddings 50/50, output head 60/40. All 427 tensors were blended with zero fallbacks; recipe in recipe.txt.

Showcase

We fused three blends of the two parents (SLERP 0.5, a 0.2→0.8 ramp, and this 0.7 one) and gave each the same briefs as Atelier on its own. This one produced the best page and the best 3D scene of the four. Unedited one-shot outputs (thinking off, temp 0.6) are in showcase/:

| | Landing page | three.js scene |

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

| Fable Atelier | !hero | !three |

| Atelier (parent) | !hero | !three |

It hasn't been benchmarked on MMLU-style tests, because it's tuned for front-end work, not trivia. The two softer blends kept more of Fable's step-by-step maths, and this one keeps Atelier's design sense.

Use

Turn thinking off for straight code output (chat_template_kwargs: {"enable_thinking": false}) and give it room, since full pages run 2-8k tokens.

llama-server -hf WineryLabs/Winery-Qwen3.5-9B-Fable-Atelier-GGUF -c 32768
ollama run hf.co/WineryLabs/Winery-Qwen3.5-9B-Fable-Atelier-GGUF

Works in llama.cpp (recent builds with Qwen3.5 support), Jan, LM Studio and the Winery app.