LLaVA-OneVision-1.5-8B-Instruct: benchmark results
Provider: LLaVA Team. Access: Open.
Aggregate Rating 1547 ± 19, rank #1732 of 10916 rated models, from 79 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| XNote (Naive RAG) | 72.84 | Naive-RAG deception-detection accuracy (%) on all 2,176 XNot | 100 |
| MCSBench - L1 | 74.71 | Accuracy (%) | 83.6 |
| XNote | 79.92 | Closed-book deception-detection accuracy (%) on all 2,176 XN | 83.3 |
| NextMotionQA - Question Answering | 52.41 | Task 1 Jaccard (%) between predicted and gold option subsets | 72.7 |
| SVLAT - Animation | 73.57 | Accuracy (%) | 70 |
| DrawingVQA - Perceptual Understanding | 61.3 | Accuracy (%; reasoning depth R1, 31 perceptual-understanding | 69.4 |
| MIBench (Multimodal Interaction) | 57.75 | Overall accuracy (%) on the full MIBench (multiple-choice im | 66.7 |
| MIBench (Multimodal Interaction) - Text-Centric | 65.38 | Accuracy (%) on the text-centric samples (the answer must co | 66.7 |
| OddGridBench - Size | 7.5 | Accuracy (%) on the 200 images in which the odd item differs | 66.7 |
| SEA-Vision - TEC-VQA (English) | 64.42 | Accuracy (%) on the English questions of SEA-Vision TEC-VQA | 63.6 |
| MIBench (Multimodal Interaction) - Synergy | 47.64 | Accuracy (%) on the synergy samples (both modalities are nee | 61.9 |
| SEA-Vision - TEC-VQA (Lao) | 12.67 | Accuracy (%) on the Lao questions of SEA-Vision TEC-VQA (tex | 59.1 |
Interactive version: theaggregate.ai/model?slug=llava-onevision-1-5-8b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.