WeEdit (Bilingual) - Replace: leaderboard
Metric: Mean instruction-adherence score (0-10 judge scale) on the replace operation of the WeEdit Bilingual Benchmark (2,000 text-centric image-editing cases), rated by a Gemini-3-Pro judge that writes a chain-of-thought rationale before scoring whether the edited image fulfils every requirement of the instruction; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nano Banana Pro | 9.06 | |
| 2 | Seedream 4.5 | 7.23 | |
| 3 | GPT-Image-1.5 | 7.16 | |
| 4 | FireRed-Image-Edit | 4.58 | |
| 5 | HunyuanImage-3.0-Instruct | 3.91 | |
| 6 | FLUX.2-dev | 3.59 | |
| 7 | Qwen-Image-Edit-2509 | 3.44 | |
| 8 | Qwen-Image-Edit-2511 | 3.36 | |
| 9 | Gemini 2.5 Flash Image | 3.35 | |
| 10 | UniWorld-V2 | 3.27 |
Interactive version: theaggregate.ai/benchmark?slug=weedit-bilingual-replace · How It Works · Data refreshed daily, snapshot 2026-10-11.