A hands-on empirical evaluation framework assessing 5 leading open-weight models (gpt-oss-20b, gpt-oss-120b, Qwen 3.8 27B, Gemma 4 31B, and Gemma 4 26B A4B) across 7 real-world task categories (including reasoning, coding, constraint compliance, and regional Indian languages). By executing a rigorous dual-blind scoring protocol under zero-drift test conditions, our project bridges the gap between synthetic leaderboard claims and practical, production-ready model selection.