Llama 3.2 Instruct 1B

Llama 3.2 Instruct 1B is a lightweight, instruction-tuned large language model developed by Meta. It is part of the Llama 3.2 collection, which specifically includes small-scale models optimized for efficiency and deployment on edge devices like smartphones and tablets.\n\n## Architecture and Training\nThe model utilizes a standard decoder-only transformer architecture with a context window of 128,000 tokens. To achieve high performance at a smaller scale, Meta developed this model through a process of pruning and knowledge distillation from the larger Llama 3.1 8B model. This technique allows the 1.23 billion parameter model to retain reasoning capabilities and knowledge while operating with significantly lower computational requirements and memory footprint.\n\n## Capabilities\nAs an instruction-tuned variant, it is optimized for conversational use cases, task-following, and multilingual dialogue. It supports eight core languages officially, including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. The model is designed for low-latency on-device applications such as summarization, rewriting, and information extraction. It is released under the Llama 3.2 Community License, supporting both research and commercial use cases.