Browse Models
The Llama 1 model family, released by Meta AI in February 2023, comprises four transformer-based language models with 7B, 13B, 33B, and 65B parameters trained on over 1 trillion publicly available tokens. The models incorporate architectural innovations including RMSNorm pre-normalization, SwiGLU activation, and rotary positional embeddings, demonstrating competitive performance across various natural language processing benchmarks while being distributed under a non-commercial research license.