mlx-community/Tmax-9B-MLX-4bit

bench: correct 27B prefill_16k numbers (false hang ruled out, hybrid attention narrative)

bench: add M3 Ultra results via rapid-mlx 0.8.18

upload 4bit

initial commit