meta-llama/Meta-Llama-3-8B-Instruct

Meta-Llama-3-8B-Instruct is an 8 billion parameter instruction-tuned generative text model developed by Meta, part of the Llama 3 family. Optimized for dialogue use cases, it utilizes an optimized transformer architecture and Grouped-Query Attention (GQA) for improved inference scalability. Trained on over 15 trillion tokens with a context length of 8192 tokens, this model is designed for commercial and research applications in English, outperforming many open-source chat models on common industry benchmarks.