File size: 1,495 Bytes
01e29bd
88c08a8
 
 
 
01e29bd
 
88c08a8
01e29bd
88c08a8
 
 
 
 
 
 
4586d4a
 
88c08a8
4586d4a
88c08a8
 
 
 
 
01e29bd
 
88c08a8
 
 
 
4586d4a
88c08a8
 
4586d4a
88c08a8
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
---
title: Qwen-Image 2.1 Rewriter Arena
emoji: ⚖️
colorFrom: indigo
colorTo: pink
sdk: gradio
sdk_version: 6.28.0
python_version: "3.12"
app_file: app.py
startup_duration_timeout: 1h
short_description: 9B vs 0.8B vs 2B prompt rewriters, rendered side by side
license: other
license_name: qwen-research
models:
  - Qwen/Qwen-Image-2.1
  - Qwen/Qwen-Image-2.1-PE-T2I
  - ML-Intern-lab/Qwen-Image-2.1-PE-T2I-Pocket-0.8B
  - ML-Intern-lab/Qwen-Image-2.1-PE-T2I-Pocket-2B
datasets:
  - ML-Intern-lab/Qwen-Image-2.1-rewriter-distill
tags:
  - text-to-image
  - prompt-rewriting
  - distillation
  - benchmark
---

# Qwen-Image 2.1 Rewriter Arena

One request goes through the official 9B prompt rewriter
([Qwen/Qwen-Image-2.1-PE-T2I](https://huggingface.co/Qwen/Qwen-Image-2.1-PE-T2I)), the two pocket distillations
([0.8B](https://huggingface.co/ML-Intern-lab/Qwen-Image-2.1-PE-T2I-Pocket-0.8B), [2B](https://huggingface.co/ML-Intern-lab/Qwen-Image-2.1-PE-T2I-Pocket-2B)),
and optionally no rewriter at all. Every rewrite is rendered by [Qwen-Image 2.1](https://huggingface.co/Qwen/Qwen-Image-2.1) with a
shared seed and shown in a grid with its prompt, token count and rewrite time. A second tab shows the 40-request evaluation
contact sheets from [ML-Intern-lab/Qwen-Image-2.1-rewriter-distill](https://huggingface.co/datasets/ML-Intern-lab/Qwen-Image-2.1-rewriter-distill)
without using any GPU.

Built with Qwen. All models are under the Qwen Research License (non-commercial research use).