VideoLLaMA3-7B: benchmark results
Provider: Alibaba. Released 2025-01-21. Access: Open.
Aggregate Rating 1528 ± 13, rank #1970 of 10916 rated models, from 269 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CLIP-CC-Bench | 67 | Mean HM-CF similarity (0-100): harmonic mean of paragraph-le | 100 |
| HAVEN - Joint Summary Keyframes | 26.7 | Keyframe selection F1 (%) against human-annotated keyframes | 100 |
| HAVEN - Keyframe Selection | 29.3 | Keyframe selection F1 (%) against human-annotated keyframes | 100 |
| ProactEgoMem - Daily IC - Notification Start@0.2 | 28.1 | Score (%) | 100 |
| ProactEgoMem - Daily LT - Notification Start@0.2 | 28.8 | Score (%) | 100 |
| ProactEgoMem - Daily LT - QA Accuracy | 36.1 | Score (%) | 100 |
| ProactEgoMem - Sci IC - Intervention Start@0.2 | 50 | Score (%) | 100 |
| ProactEgoMem - Sci LT - Intervention Recall | 50.9 | Score (%) | 100 |
| ProactEgoMem - Sci LT - Notification Start@0.2 | 40.3 | Score (%) | 100 |
| ProactEgoMem - Sci LT - Notification mIoU | 14.6 | Score (%) | 100 |
| ProactEgoMem - Sci ST - Notification Start@0.2 | 30 | Score (%) | 100 |
| TOC-Bench - Event Count | 33.7 | Event Count accuracy (%) on TOC-Bench (2,323 human-verified | 100 |
Interactive version: theaggregate.ai/model?slug=videollama3-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.