Add detailed model card with training details and GSM8K results

Upload CoT-SFT fine-tuned model (Qwen3-0.6b-thinking)

initial commit