TinyGrafixBench - Network: leaderboard
TinyGrafixBench mIoU (%) on network edits (adding, swapping, removing and recoloring nodes), scored pixel by pixel against the single correct image; higher is better (PaintBench paper, NYU, 2026).
Metric: mIoU (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nano-Banana-2 | 4.8 | |
| 2 | GPT-Image-2 | 4.8 | |
| 3 | Nano Banana | 2.7 | |
| 4 | FLUX.2-klein-9B | 2.4 | |
| 5 | FLUX.2-dev | 1.9 | |
| 6 | FLUX.1-Kontext-dev | 1.9 | |
| 7 | BAGEL-7B-MoT | 1.8 | #5515 |
| 8 | Qwen-Image-Edit-2511 | 1.6 | |
| 9 | LongCat-Image-Edit | 0.8 | |
| 10 | HunyuanImage-3.0 | 0.1 |
Interactive version: theaggregate.ai/benchmark?slug=tinygrafixbench-network · How It Works · Data refreshed daily, snapshot 2026-11-12.