LLaVA-NeXT-Video-34B: benchmark results
Provider: LLaVA Team. Access: Open.
Aggregate Rating 1448 ± 27, rank #3878 of 10916 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM Video - MMBench-Video - Hallucination | 0.58 | GPT-4 Rating (0-3) | 58 |
| Flat-Pack Bench - Part Tracking | 32.68 | Track: recover the correspondence of shuffled part IDs betwe | 50 |
| LongVideoBench | 50.5 | Test Total Accuracy (%) | 50 |
| OpenVLM Video - MMBench-Video - Commonsense Reasoning | 1.27 | GPT-4 Rating (0-3) | 36.4 |
| OpenVLM Video - MMBench-Video - Relation Reasoning | 1.3 | GPT-4 Rating (0-3) | 35.2 |
| OpenVLM Video - MMBench-Video - Attribute Reasoning | 1.38 | GPT-4 Rating (0-3) | 31.8 |
| OpenVLM Video - MMBench-Video - Cross-Instance Fine-grained Perception | 0.97 | GPT-4 Rating (0-3) | 31.8 |
| OpenVLM Video - MMBench-Video - Single-Instance Fine-grained Perception | 1.15 | GPT-4 Rating (0-3) | 29.5 |
| OpenVLM Video - MMBench-Video - Coarse Perception | 1.35 | GPT-4 Rating (0-3) | 27.3 |
| OpenVLM Video - MMBench-Video - Temporal Reasoning | 1.03 | GPT-4 Rating (0-3) | 27.3 |
| OpenVLM Video - MMBench-Video | 38 | Normalized Score (%) | 27.1 |
| Vinoground (Text Score) | 23 | Text Score | 26.1 |
Interactive version: theaggregate.ai/model?slug=llava-next-video-34b · How It Works · Data refreshed daily, snapshot 2026-11-08.