I am creating a RAG application based on legal texts. I have documents approx. 80 million tokens long. As you know for RAG application I need to embed documents first.
Formerly I was using openai embeddings text-embedding-3-large model but it will cost 10k$. For better outputs and lower cost I want to switch Nvidia nim nv-embed-v1 but I have never hosted an AI model before so I cannot predict approx. cost.
I will be glad if you guys can share any source for hosting AI models, calculating Host Cost and maybe tell me which one is cheaper text-embedding-3-large or Nvidia nim nv-embed-v1?
Thank you.