I tried different models but I am stuck in this problem for almost a week. I am working on a QA model using the Retrieval-Augmented Generation (RAG) technique. I create a vector database of all input articles and use FAISS to find the closest 3 articles to create context. I then send the question and context in a prompt to GPT-3.5-turbo to generate a fully developed answer based on the context. GPT-3.5-turbo generates excellent and accurate results, but it is costly for me. Therefore, I tried GPT-2-xl and GPT-Neo, but they are not even close to GPT-3.5-turbo and most of the time they generate distant results from the context. Is there any Free model available that is almost similar to GPT-3.5-turbo (generation and accuracy)? [I want to generate a fully developed answer like this for the question “How much worth of products is the ban on Russian steel imports estimated to affect?”: “The ban on Russian steel imports is estimated to affect 3.3 billion euros ($3.6 billion) worth of products. This figure was stated by the European Commission as part of a larger package of sanctions against Russia in response to its invasion of Ukraine,” not just “$3.6 billion.”] Thanks a lot for the help.