What Qwen3-Coder model are you using? Quantized or not?

Asking because I'm looking for a good model that fits in 12GB VRAM.