LLaVA-v1.5-7B: benchmark results
Provider: LLaVA Team. Released 2023-10-05. Access: Open.
Aggregate Rating 1342 ± 18, rank #7935 of 10916 rated models, from 674 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| GeoHeight-Bench - Pattern Description | 27.59 | Text accuracy (%) on the Pattern Description task (the model | 100 |
| OpenVLM MMT-Bench - VI | 68.1 | Score (%) | 96.6 |
| OpenVLM MMT-Bench - Animal Keypoint Detection | 50 | Score (%) | 96.4 |
| Open LMM Reasoning - WeMath - RoteMemorization (Loose) | 23.3 | Accuracy (%) | 96.3 |
| Open LMM Reasoning - WeMath - RoteMemorization (Strict) | 83.6 | Accuracy (%) | 96.3 |
| OpenVLM MMT-Bench - Geometrical Relativity | 75 | Score (%) | 96.1 |
| MIR-SafetyBench - Semantic Relevance | 20.39 | Attack Success Rate (%; lower is safer) | 94.4 |
| MIR-SafetyBench - Spatial Juxtaposition | 40.07 | Attack Success Rate (%; lower is safer) | 94.4 |
| MIR-SafetyBench - Temporal Continuity | 34.7 | Attack Success Rate (%; lower is safer) | 94.4 |
| MIR-SafetyBench - Temporal Jump | 36.96 | Attack Success Rate (%; lower is safer) | 94.4 |
| OpenVLM MMT-Bench - Deepfake Detection | 65 | Score (%) | 93.7 |
| Open LMM Reasoning - WeMath - InsufficientKnowledge (Loose) | 67.4 | Accuracy (%) | 93.2 |
Interactive version: theaggregate.ai/model?slug=llava-v1-5-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.