Imagine coding even complex projects right on your laptop, without needing powerful hardware. Recent advances in AI are making this possible, bringing the power of large language models (LLMs) to everyday devices through a technique called quantization. LLMs have shown incredible potential for tasks like generating and understanding code, but their size often demands substantial resources, limiting their use to high-end machines. Quantization changes this by compressing these large models, making them leaner and faster while retaining much of their ability to generate working code. This research delves into the viability of quantized LLMs for code generation, focusing on how well they perform on less common programming languages like Lua, often used in specialized applications like game scripting and embedded systems. Why Lua? Because it's different from mainstream languages, and testing on it gives us a more realistic picture of how these compressed models handle diverse coding tasks. The study investigates different levels of quantization—think of it like adjusting image quality—to find the sweet spot between model size and performance. Interestingly, the results show that 4-bit quantization provides an excellent balance, allowing models with billions of parameters to run efficiently on ordinary laptops. These quantized models were even better than smaller, uncompressed models at generating correct Lua code. While there's still work to do, this research highlights a promising path to democratizing AI coding assistants, enabling anyone with a regular computer to harness the power of these advanced tools.