Imagine combining the strengths of different AI minds without any extra training. That's the magic of Cool-Fusion, a new technique that blends the knowledge of multiple Large Language Models (LLMs) to create a smarter, more versatile AI. LLMs, despite their impressive abilities, each have unique strengths and weaknesses. Cool-Fusion leverages these differences by having each LLM independently generate text segments. Then, like a team of experts evaluating different solutions, the LLMs collectively assess and rank the generated text, selecting the best option based on a combined "perplexity" score – a measure of how uncertain the model is about its prediction. This approach eliminates the need for costly retraining or vocabulary alignment, making it a faster and more efficient way to boost LLM performance. The research team used several state-of-the-art LLMs with distinct vocabularies to test this approach. Results show that Cool-Fusion significantly outperforms individual LLMs on challenging tasks like math word problems, question answering, and multilingual understanding. In essence, this method picks the best parts from each LLM and merges them into one high-quality solution. This research unlocks a simple yet powerful way to enhance AI capabilities, which could revolutionize how we build and use AI in the future. By harnessing the complementary strengths of multiple LLMs, Cool-Fusion offers a glimpse into a future where collaboration between AI minds leads to extraordinary advancements in knowledge and problem-solving.