ChronoVision - News Photo Year: leaderboard
News photo dating on ChronoVision (2026), a temporal reasoning benchmark for vision-language models: the percent of 1,028 dated event photos from 1952 to 2025 whose year the model predicts correctly. Higher is better.
Metric: Year accuracy (%) on the 1,028 dated event photos (1952-2025): predict the year a photo was taken; zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 55.58 | #266 |
| 2 | Qwen 3 VL 235B A22B Instruct | 48.48 | #522 |
| 3 | GPT-5.2 | 40.98 | #176 |
| 4 | Qwen 3 VL 8B Instruct | 35.8 | #981 |
| 5 | GLM-4.1V-9B (Thinking) | 34.26 | #1060 (GLM-4.1V-9B) |
| 6 | Qwen 2.5 VL 7B Instruct | 34.12 | #1833 |
| 7 | Qwen 3 VL 4B Instruct | 32.86 | #1279 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chronovision-news-photo-year · How It Works · Data refreshed daily, snapshot 2026-11-13.