A 7-billion parameter instruction-tuned LLM from Mistral AI optimized for efficient text generation and reasoning.