Llama 3.2
About
Llama 3.2, released by Meta, represents an advanced family of multilingual large language models (LLMs) that cater to both text-only and multimodal applications. Designed for flexibility, these models range in size to accommodate various computational capabilities 1. The smaller models, with parameters of 1B and 3B, are optimized for on-device functions such as summarization and instruction following, featuring a remarkable context length of 128K tokens, establishing them as leaders in their category 1. Conversely, the larger models, ranging up to 90B parameters, are enabled with vision capabilities, allowing them to undertake tasks like image captioning and visual reasoning 1. Supporting a diverse array of languages, including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai, these models are versatile in their application, easily accessible via Meta's Llama website and the Hugging Face platform 3.