Hourly Usage
Performance Metrics
Median Total Time
22.01s
Median TTFT
1.25s
Median Prefill TPS
1789.55
Median Gen TPS
34.43
Model Information
Context Size
262144
Quantization
r128 on INT8
Engine
vllm
Creation Method
Finetune
Model Type
Qwen38
Chat Template
Qwen3.5
Reasoning
Yes
Vision
Yes
Parameters
27B
Added At
9/28/2026
thumbnail: >- https://cdn-uploads.huggingface.co/production/uploads/634262af8d8089ebaefd410e/LzXxeS6Rkun66x_UEoDKi.jpeg license: apache-2.0 language:
- en base_model:
- Qwen/Qwen3.8-27B tags:
- not-for-all-audiences
- roleplay
- creative-writing
- axolotl
- unsloth
- slop? peak? you decide!
- using huggingface tags like AO3
- no beta read we die like llama 4
The newest member of your roleplay stack
Qwen 3.8 27B Dominatrix
A roleplay-focused finetune for improved prose and more creative reasoning.
Description
An RP-focused finetune of Qwen 3.8 27B for improved prose and better (or at least more creative) reasoning.
Findom
Here's some links to support Fizz, the member who trained this model:
Support Fizz on Ko-fiAs well as her Monero address:
Chat Template Changes
The original chat template has had its defaults adjusted. If you notice anything, it was intentional.
Usage
Chat format: standard Qwen3.8-style ChatML. A chat template mismatch is the one thing that visibly disappoints her.
Samplers: the author ran her comfortably at temperature 1.0 to 1.25 with either min_p 0.1 or top_p 0.95. Others found success at 0.7 temp and nothing else. Iunno!
Acknowledgements
- Everyone at Allura, for being cool and based.
- Everyone whose data this model was trained on, and everyone who collated it.
- The global proletariat.