Qwen-Image-Bench - Human Expert Rating - 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 professional annotators rating of the real-world fidelity (knowledge-consistent depiction of real entities, text and physical facts) pillar on a 1 to 10 scale; higher is better.
Metric: Real-world fidelity (knowledge-consistent depiction of real entities, text and physical facts) pillar professional annotator rating on a 1 to 10 scale over 1,000 prompts per pillar, averaged; 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 | 8.8 | |
| 2 | Nano Banana 2 | 8.64 | |
| 3 | Nano Banana Pro | 8.45 | |
| 4 | GPT Image 1.5 | 7.29 | |
| 5 | Qwen Image 2.0 Pro | 7.27 | |
| 6 | Seedream 5.0 | 6.22 | |
| 7 | Qwen-Image-2512 | 5.23 | |
| 8 | Seedream 4.5 | 5.17 | |
| 9 | GPT Image 1 | 4.98 | |
| 10 | Seedream 4.0 | 4.67 | |
| 11 | Imagen 4.0 Ultra | 4.63 | |
| 12 | Imagen 4.0 | 4.51 | |
| 13 | FLUX 2 Max | 4.07 | |
| 14 | FLUX 2 Pro | 3.41 | |
| 15 | HunyuanImage 3.0 | 3.08 |
Interactive version: theaggregate.ai/benchmark?slug=qwen-image-bench-human-expert-rating-real-world-fidelity · How It Works · Data refreshed daily, snapshot 2026-11-21.