A 70-billion-parameter base model offering strong performance in natural language processing tasks. It serves as a foundation for various fine-tuned versions tailored to specific applications. Supports a context length of 8K tokens.