What is currently the best LLM model for consumer grade hardware? Is it phi-4?

I have a 5060ti with 16GB VRAM. I’m looking for a model that can hold basic conversations, no physics or advanced math required. Ideally something that can run reasonably fast, near real time.