After conducting several research considering models we have already used, our next top choices were Mistral, Falcon, Phi, and Claude. From these, Mistral 7B had surpassed several models' benchmarks including Llama 2's 13B on all benchmarks. Adding on, there were several quantized versions for this LLM with different loss and sizes on hugging face, hence giving us the flexibility to choose the best model for our scenario.
After conducting several research considering models we have already used, our next top choices were Mistral, Falcon, Phi, and Claude. From these, Mistral 7B had surpassed several models' benchmarks including Llama 2's 13B on all benchmarks. Adding on, there were several quantized versions for this LLM with different loss and sizes on hugging face, hence giving us the flexibility to choose the best model for our scenario.