MVX-Bench - Video Similarity: leaderboard
Metric: Accuracy (%) on the 69 Video Similarity questions (pick the near-duplicate of the reference video), MVX-Bench's multi-video multiple-choice questions (each pairs a question with several videos; 1,442 questions over 4,255 videos repurposed from 11 annotated video datasets), end-to-end inference at temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 27B (IT) | 82.6 | #509 |
| 2 | GPT-5.2 | 81.2 | #105 |
| 3 | Qwen 2.5 VL 32B Instruct | 78.3 | #443 |
| 4 | GPT-4o | 76.8 | #333 |
| 5 | Gemma 3 4B (IT) | 72.5 | #971 |
| 6 | Gemma 3 12B (IT) | 71 | #655 |
| 7 | Qwen 3.6 27B (Thinking) | 53.6 | #163 (Qwen 3.6 27B) |
| 8 | Pixtral-12B | 36.2 | #795 |
| 9 | Qwen 3.6 27B (Non-reasoning) | 26.1 | #163 (Qwen 3.6 27B) |
| 10 | Qwen 2.5 VL 7B Instruct | 23.2 | #643 |
| 11 | Phi-4 Multimodal Instruct | 21.7 | #896 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=mvx-bench-video-similarity · How It Works · Data refreshed daily, snapshot 2026-10-11.