We have used Mistral 7B chat model and it looks promising in few use cases mentioned in above comments.

Unfortunately, I don't have very specific numbers compared to other models like PaLM or GPT3.5/4 but based on multiple manual evaluation we found Mistral performs really well against PaLM model.

One specific example where Mistral was able to read and interpret Chinese language where PaLM failed to identify.

Accuracy at the end will largely depend on the data and use case.