Llama 3.2 90B Vision Instruct — benchmark results
Meta Llama 3.2 90B vision instruction-tuned checkpoint. Provider: Meta. Released 2024-09-25. Access: Open.
Unified ELO 1478 ± 8, rank #879 of 1776 rated models, from 450 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM SeaHELM - IndicQA | 57.23 | SQuAD macro-averaged F1 score | 100 |
| OpenVLM MMT-Bench - Face Detection | 95 | Score (%) | 100 |
| OpenVLM MMT-Bench - Writing Poetry From Image | 85 | Score (%) | 99.8 |
| OpenVLM MMBench V1.1 EN - Function Reasoning | 97.8 | Accuracy (%) | 99.5 |
| OpenVLM MMT-Bench - Film and Television Recognition | 100 | Score (%) | 99 |
| OpenVLM MMT-Bench - Traffic Anomaly Detection | 70 | Score (%) | 99 |
| OpenVLM MMT-Bench - Plant Recognition | 95 | Score (%) | 98.1 |
| OpenVLM MMT-Bench - VR | 85.3 | Score (%) | 97.8 |
| OpenVLM MMT-Bench - Abstract Visual Recognition | 90 | Score (%) | 96.8 |
| OpenVLM MMT-Bench - Sign Language Recognition | 50 | Score (%) | 96.8 |
| OpenVLM MMT-Bench - AR | 58 | Score (%) | 96.6 |
| OpenVLM MMT-Bench - GUI General | 65 | Score (%) | 96.6 |
Interactive version: theaggregate.ai/model?slug=llama-3-2-90b-vision-instruct · How the rankings work · Data refreshed daily, snapshot 2026-07-22.