Vinoground (Group Score): leaderboard
Vinoground temporal counterfactual benchmark, group score requiring both text and video matches correct for a pair to count.
Metric: Group Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 24.6 | #343 |
| 2 | Qwen 2 VL 72B | 17.4 | #460 |
| 3 | Qwen 2 VL 7B | 15.2 | #750 |
| 4 | Claude 3.5 Sonnet | 10.6 | #350 |
| 5 | Gemini 1.5 Pro | 10.2 | #445 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vinoground-group-score · How It Works · Data refreshed daily, snapshot 2026-10-20.