GeoR-Bench - Consistency: leaderboard
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. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Nano Banana Pro | 96.3 |
| 2 | GPT-Image-2 | 95.1 |
| 3 | Nano Banana 2 | 94.6 |
| 4 | GPT-Image-1.5 | 94.4 |
| 5 | Seedream 5.0 | 90.3 |
| 6 | Nano Banana | 88.1 |
| 7 | FLUX 2 Max | 71.6 |
| 8 | Seedream 4.5 | 71.3 |
| 9 | Qwen-Image-Edit-2511 | 62.6 |
| 10 | FLUX.2-dev | 62.2 |
Interactive version: theaggregate.ai/benchmark?slug=geor-bench-consistency · How It Works · Data refreshed daily, snapshot 2026-10-07.