DOLPHIN 3.0 (HUGGING FACE HUB)

Summary from the TLDR AI newsletter

The new Dolphin model, built on Llama 3.1 8B, was trained to take instructions via the system prompt. This means that it is exceptionally good at various personas and roleplay while maintaining much of the power of the base Llama model itself.