LLM Stats (MLVU): leaderboard
Multi-task Long Video Understanding: evaluating comprehension of long videos across topic reasoning, anomaly detection, and needle-in-a-haystack retrieval.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Plus | 87.4 | #92 |
| 2 | Qwen 3.5 122B A10B | 87.3 | #173 |
| 3 | Qwen 3.6 Plus | 86.7 | #120 |
| 4 | Qwen 3.6 27B | 86.6 | #170 |
| 5 | Qwen 3.6 35B A3B | 86.2 | #226 |
| 6 | Qwen 3.5 27B | 85.9 | #187 |
| 7 | Qwen 3.5 35B A3B | 85.6 | #256 |
| 8 | Qwen 3 VL 235B A22B Instruct | 84.3 | #273 |
| 9 | Qwen 3 VL 235B A22B (Thinking) | 83.8 | #238 (Qwen 3 VL 235B A22B) |
| 10 | Qwen 2.5 VL 7B Instruct | 70.2 | #654 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-mlvu · How It Works · Data refreshed daily, snapshot 2026-10-20.