GeoR-Bench - Consistency: leaderboard

Consistency score on a 0-100 scale: whether the edited image keeps the visual relationship to the input that the task requires, averaged over 440 geoscience image-editing samples in six categories (GeoR-Bench, 2026), judged by Gemini 3 Flash with task-specific rubrics. Higher is better.

Metric: Consistency score (0-100): whether the output preserves the visual relationship to the input image that the task requires (mean of the six category scores), 440 reasoning-informed geoscience image-editing samples in six categories (geomorphology, hydrology, atmosphere and ocean, cryosphere, GIS and spatial geometry, crustal science), one generated output per sample, Gemini 3 Flash judge with task-specific rubrics; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Nano Banana Pro96.3
2GPT Image 295.1
3Nano Banana 294.6
4GPT Image 1.594.4
5Seedream 5.090.3
6Nano Banana88.1
7FLUX 2 Max71.6
8Seedream 4.571.3
9Qwen-Image-Edit-251162.6
10FLUX.2-dev62.2
11FLUX 2 Pro61.3
12Seedream 4.058.3
13DreamOmni57.6
14Step1X-Edit50.4
15BAGEL-7B-MoTThinking44.5#4668

Interactive version: theaggregate.ai/benchmark?slug=geor-bench-consistency · How It Works · Data refreshed daily, snapshot 2026-11-22.