VerifAI's MultiLLM is an open-source Python framework that allows users to leverage the power of multiple Language Model Models (LLMs) simultaneously. By invoking multiple LLMs in parallel and ranking their outputs, VerifAI's MultiLLM aims to find the most accurate results, also known as the ground truth. The initial use case for MultiLLM focuses on comparing code generated by popular LLMs such as GPT3, GPT5, and Google-Bard. However, this framework can be extended to support new LLMs and enables the customization of ranking functions to evaluate a diverse range of outputs from different LLMs. With its flexible and versatile nature, VerifAI's MultiLLM offers users the ability to obtain reliable results for various tasks. Whether users need to request code or seek answers to specific questions, MultiLLM utilizes multiple LLMs simultaneously and ranks their responses to provide the most accurate and best-performing outcomes. It is worth noting that an individual LLM may occasionally provide incorrect information about people, places, or facts. Therefore, by combining the outputs of multiple LLMs and comparing their results using VerifAI's MultiLLM framework, users can mitigate the risk of relying solely on potentially erroneous information. For those interested in exploring further, the MultiLLM framework is open-source and available on GitHub, and additional information can be found in the associated VerifAI blog article.
