Update model card with precise metrics, evaluation transparency, and context engineering examples

Fine-tuned Llama 3.1 8B adapter for structured API generation - 40% accuracy (95% better than Azure GPT)

initial commit