Salesforce produced one of the best Llama-3(8B) finetunes, IMO: SFR-Iterative-DPO-LLaMA-3-8B-R

Hopefully they do something with Llama-3.1