CHOICE: leaderboard
Remote-sensing vision-language benchmark for perception and reasoning over Earth-observation imagery across multiple task dimensions.
Metric: Overall Score. Source: rs-choice.github.io. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 VL 72B | 0.74 | #460 |
| 2 | Gemini 1.5 Pro | 0.69 | #445 |
| 3 | Qwen 2 VL 7B | 0.69 | #750 |
| 4 | InternVL2-8B | 0.68 | #817 |
| 5 | GPT-4o (2024-11-20) | 0.63 | #381 |
| 6 | GPT-4o Mini | 0.61 | #599 |
| 7 | Llama 3.2 11B | 0.57 | #1199 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=choice · How It Works · Data refreshed daily, snapshot 2026-10-20.