ChronoBench - Spatio-Temporal Reasoning: leaderboard
Metric: Accuracy (%; level average of class-level change magnitude estimation, region development comparison and object construction ordering within and across sequences; 17,689 validated multiple-choice questions over remote-sensing image time series, box- and geo-coordinate-grounded variants; open models greedy with 128 new tokens, API models at high reasoning effort). Source: arxiv.org. Saturation forecast: Around 2029. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 59.89 |
| 2 | GPT-5.4 (High) | 50.42 |
| 3 | Qwen 3 VL 32B Instruct | 47.23 |
| 4 | InternVL3.5-8B | 45.11 |
| 5 | Qwen 3 VL 8B Instruct | 42.8 |
| 6 | Qwen 3 VL 4B Instruct | 39.53 |
Interactive version: theaggregate.ai/benchmark?slug=chronobench-spatio-temporal-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-29.