TVBench: leaderboard
Video understanding benchmark reported in Qwen's Qwen3.7-Plus comparison table.
Metric: Average Accuracy. Source: qwen.ai. Source: huggingface.co. Saturation forecast: Estimated already saturated. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 VL 72B | 52.7 | #460 |
| 2 | Gemini 1.5 Pro | 47.6 | #445 |
| 3 | Qwen 2 VL 7B | 43.8 | #750 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=tvbench · How It Works · Data refreshed daily, snapshot 2026-10-20.