ChronoVision - Artifact Dating: leaderboard
Artifact dating accuracy on ChronoVision (2026), a temporal reasoning benchmark for vision-language models: the percent of 887 Chinese historical artifact images, from six crafts spanning the Tang to the Qing dynasty, whose dynasty the model names correctly. Higher is better.
Metric: Accuracy (%) on the 887 Chinese historical artifact images (six crafts, Tang to Qing): choose the dynasty; zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B Instruct | 49.94 | #522 |
| 2 | Gemini 2.5 Pro | 44.23 | #266 |
| 3 | Qwen 2.5 VL 7B Instruct | 44.01 | #1833 |
| 4 | GPT-5.2 | 40.14 | #176 |
| 5 | Qwen 3 VL 4B Instruct | 40.14 | #1279 |
| 6 | Qwen 3 VL 8B Instruct | 38.43 | #981 |
| 7 | GLM-4.1V-9B (Thinking) | 36.53 | #1060 (GLM-4.1V-9B) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chronovision-artifact-dating · How It Works · Data refreshed daily, snapshot 2026-11-13.