Best LLMWare model for RAG

[Preet1234-ux]

I have used the model "llmware/bling-phi-3-gguf" for my RAG pipeline . But the accuracy is not good . I have got 66% accuracy for my dataset using this model . I have also seen that if we increase the context retrieval size , the accuracy for some queries was reduced .
Can you suggest few best models for RAG ?

[scientific-coder]

This could be an opportunity to share models that could be interesting to investigate. What about domain specific models for instance ?
I recently discovered SaulLM with models for legal documents.
What is your dataset about ?

[Gokul-MK]

Several models stand out for boosting accuracy in your RAG pipeline. GPT-4-turbo from OpenAI is highly effective for handling large contexts, accommodating up to 32K tokens, making it ideal for nuanced retrieval tasks. Claude 2 by Anthropic, with its capacity for up to 100K tokens, is excellent for managing broader contexts without compromising accuracy. FLAN-T5 by Google is efficient and accurate, particularly when fine-tuned, and provides a good balance for smaller setups.