LLM-MS: A Multi-Model LLM Search Engine

Large Language Models (LLMs) are evolving at an unprecedented pace, with new models emerging almost daily. This has created a significant challenge for users trying to identify the most suitable model for a given task. With thousands of models available, selecting the right LLM has become increasingly complex. Retraining or fine-tuning a single model for general-purpose use is computationally expensive, making specialized models a practical necessity. However, the lack of tools to dynamically evaluate and combine these models limits their accessibility and effectiveness. In this paper, we introduce LLM Meta Search (LLM-MS), a web-based multi-model system that dynamically selects and allocates resources to LLMs based on query characteristics. To achieve this, we propose novel LLM selection algorithms that optimize token allocation, balancing response accuracy and computational cost. Our evaluation demonstrates that LLM-MS significantly reduces token usage and response time compared to traditional single-model approaches while maintaining high-quality outputs.