Qwen-Image-Bench - Real-World Fidelity: leaderboard
Qwen-Image-Bench (2026) scores text-to-image models on 1,000 creator-designed prompts in five pillars: quality, aesthetics, alignment, real-world fidelity and creative generation. This board is the real-world fidelity (knowledge-consistent depiction of real entities, text and physical facts) pillar score from the judge model on the Chinese prompts, 0 to 100; higher is better.
Metric: Real-world fidelity (knowledge-consistent depiction of real entities, text and physical facts) pillar Q-Judger score (0-100), the unified judge model fine-tuned from Qwen3.6-27B on expert annotations, aggregated bottom-up from 56 third-level facets, over the 1,000 creator-designed Chinese prompts; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 18 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT Image 2 | 57.38 | |
| 2 | Nano Banana 2 | 54.28 | |
| 3 | Nano Banana Pro | 54.07 | |
| 4 | GPT Image 1.5 | 53.95 | |
| 5 | Seedream 5.0 | 51.92 | |
| 6 | Qwen Image 2.0 Pro | 51.83 | |
| 7 | Seedream 4.5 | 51.69 | |
| 8 | Seedream 4.0 | 51.05 | |
| 9 | FLUX 2 Max | 49.35 | |
| 10 | GPT Image 1 | 48.14 | |
| 11 | FLUX 2 Pro | 47.29 | |
| 12 | Qwen-Image-2512 | 47 | |
| 13 | Imagen 4.0 Ultra | 45.59 | |
| 14 | Imagen 4.0 | 44.84 | |
| 15 | Kling Image 2.1 | 44.74 |
Interactive version: theaggregate.ai/benchmark?slug=qwen-image-bench-real-world-fidelity · How It Works · Data refreshed daily, snapshot 2026-11-21.