InEdit-Bench: leaderboard
Metric: Overall average (0-100) of six evaluator-scored dimensions (appearance consistency, perceptual quality, semantic consistency, logical coherence, scientific plausibility, process plausibility), where the editing model generates the intermediate stages of a multi-step transformation from an initial image to a described final state, scored 0-100 by a GPT-4o-2024-11-20 evaluator; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-Image-1 | 81.33 | |
| 2 | Gemini 2.5 Flash Image (Preview) (Nano Banana) | 75.23 | |
| 3 | Flux-Kontext-pro | 51.46 | |
| 4 | Qwen-Image-Edit | 49.6 | |
| 5 | BAGEL-7B-MoT | 40.7 | |
| 6 | BAGEL-7B-MoT (Thinking) | 40.7 | |
| 7 | OmniGen2 | 37.92 | |
| 8 | SeedEdit 3.0 | 36.5 | |
| 9 | Step1X-Edit-v1p1 | 31.39 | |
| 10 | Emu2 (InEdit-Bench checkpoint unspecified) | 29.61 |
Interactive version: theaggregate.ai/benchmark?slug=inedit-bench · How It Works · Data refreshed daily, snapshot 2026-10-11.