meta-llama/Llama-3.1-8B-Instruct

Llama 3.1 8B Instruct is an 8 billion parameter instruction-tuned generative language model developed by Meta, optimized for multilingual dialogue and general natural language generation tasks. It features a 128K token context length and utilizes an optimized transformer architecture with Grouped-Query Attention for improved inference scalability. Trained on over 15 trillion tokens with a December 2023 knowledge cutoff, this model excels in areas like reasoning, code generation, and tool use, outperforming its predecessor Llama 3 8B Instruct on various benchmarks.