Llama 3.2: Revolutionizing edge AI and vision with open, customizable models

Benchmark Model Rank Results
math-word-problem-solving-on-mgsm-0-shotLlama 3.2 90B Vision Instruct#4Accuracy: 86.9
math-word-problem-solving-on-mgsm-0-shotLlama 3.2 3B Instruct#7Accuracy: 58.2
math-word-problem-solving-on-mathLlama 3.2 90B Vision Instruct#37Accuracy: 68.0
math-word-problem-solving-on-mathLlama 3.2 3B Instruct#94Accuracy: 48.0
chart-question-answering-on-chartqaLlama 3.2 90B Vision Instruct#411:1 Accuracy: 85.5
visual-question-answering-on-docvqa-testLlama 3.2 90B Vision Instruct#44ANLS: 0.901
multi-task-language-understanding-on-mmluLlama 3.2 3B Instruct#45Average (%): 63.4
multimodal-reasoning-on-mmmuLlama 3.2 90B Vision Instruct#58Accuracy: 60.3
multimodal-reasoning-on-mathvistaLlama 3.2 90B Vision Instruct#100Accuracy: 57.3
sentence-completion-on-hellaswagLlama 3.2 3B Instruct#111Accuracy: 69.8
arithmetic-reasoning-on-gsm8kLlama 3.2 3B Instruct#166Accuracy: 77.7