LLaMA-VID-7B: benchmark results
Provider: Meta.
Aggregate Rating 1433 ± 36, rank #4514 of 10916 rated models, from 88 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MLVU-Dev - Generation | 4.22 | Generation score, GPT-4 judged (0-10) | 79.3 |
| OpenVLM Video - MLVU - Action Count | 33.5 | Accuracy (%) | 56.9 |
| OpenVLM Video - MVBench - Egocentric Navigation | 34.5 | Accuracy (%) | 51.4 |
| OpenVLM Video - MLVU - Ego Reasoning | 51.7 | Accuracy (%) | 47.2 |
| OVBench | 41.9 | AVG | 46.7 |
| OpenVLM Video - MLVU - Generation Score | 3.95 | Judge Score (0-10) | 45.7 |
| MLVU-Test - Action Order | 18.6 | Multiple-choice accuracy (%) | 42.6 |
| OpenVLM Video - MMBench-Video - Hallucination | 0.42 | GPT-4 Rating (0-3) | 42 |
| MLVU-Test - Action Count | 15 | Multiple-choice accuracy (%) | 38.2 |
| MLVU-Test - Needle QA | 21.7 | Multiple-choice accuracy (%) | 38.2 |
| OpenVLM Video - MVBench - Scene Transition | 88 | Accuracy (%) | 36.1 |
| OpenVLM Video - MVBench - Action Count | 42 | Accuracy (%) | 34.7 |
Interactive version: theaggregate.ai/model?slug=llama-vid-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.