ChronoBench: leaderboard
ChronoBench (2026) tests whether multimodal LLMs track, remember and reason about land-cover change across long remote-sensing image time series, with 12 sub-tasks at four cognitive levels. Overall accuracy in percent over all questions, higher is better.
Metric: Overall accuracy (%; all 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 July 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (High) | 56.29 | #123 (GPT-5.4) |
| 2 | Qwen 3 VL 32B Instruct | 46.73 | #585 |
| 3 | Qwen 3 VL 8B Instruct | 39.19 | #981 |
| 4 | Qwen 3 VL 4B Instruct | 35.1 | #1279 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chronobench · How It Works · Data refreshed daily, snapshot 2026-11-13.