ChronoPhyBench - Frame Sorting (3 Frames): leaderboard
Metric: Accuracy (%) on ChronoPhyBench chronological sorting of 3 held-out frames: given the video context and caption, put the frames in the order the physical process produces them (exact order required); chance 16.7; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Doubao-Seed-2.0-Pro-260215 | 74 |
| 2 | GPT-5.5 | 72 |
| 3 | Gemini 3 Pro (Preview) | 70 |
| 4 | Doubao-Seed-1-8-251228 | 61 |
| 5 | GPT-5.3 | 53 |
| 6 | Doubao-Seed-1.6-thinking-250715 | 42 |
| 7 | Qwen 3 VL 32B Instruct | 41 |
| 8 | GPT-5.4 | 32.5 |
| 9 | Qwen 3 VL 8B Instruct | 31.5 |
| 10 | InternVL3-8B | 30.5 |
| 11 | GLM-4.1V-9B (Thinking) | 29.5 |
| 12 | InternVL3-38B | 23.5 |
| 13 | InternVL3-14B | 23.5 |
| 14 | Qwen 3 VL 8B (Thinking) | 23.5 |
| 15 | InternVL3.5-8B | 20 |
Interactive version: theaggregate.ai/benchmark?slug=chronophybench-frame-sorting-3-frames · How It Works · Data refreshed daily, snapshot 2026-09-29.