VideoChat-Flash-7B: benchmark results
Provider: Other.
Aggregate Rating 1525 ± 39, rank #2011 of 10916 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MLVU-Dev | 74.7 | Multiple-choice accuracy (%) | 94.7 |
| TimeBlind - Accuracy | 64.9 | Standard accuracy (%) over all 2,400 TimeBlind video-questio | 61.9 |
| TimeBlind | 19.5 | Instance accuracy I-Acc (%) over TimeBlind's 600 instances ( | 47.6 |
| VRBench | 50.82 | Overall (imported) | 35.7 |
| LongEgoRefer - Temporal Grounding | 1.07 | Mean temporal IoU (%; predicted versus annotated time interv | 33.3 |
| Molmo2 Human Preference | 956 | Elo rating (bootstrap median, Bradley-Terry; pairwise human | 25 |
| VirtueBench (256 Frames) | 27.12 | Accuracy (%) over VirtueBench's 1,180 open-ended long-video | 20.5 |
| Molmo2 Human Preference - Captioning | 932 | Elo rating (bootstrap median, Bradley-Terry; pairwise human | 20 |
| Molmo2-CapTest | 14.8 | Caption F1 (%; LLM-judged statement precision and recall aga | 20 |
| Molmo2-VideoCount | 21.6 | Exact-count accuracy (%; 533 validation queries) | 5 |
| VirtueBench (128 Frames) | 24.35 | Accuracy (%) over VirtueBench's 1,191 open-ended long-video | 2.1 |
| VirtueBench (1024 Frames) | 22.66 | Accuracy (%) over VirtueBench's 587 open-ended long-video qu | 0 |
Interactive version: theaggregate.ai/model?slug=videochat-flash-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.