How can I evaluate which model performs best for my dataset?

Benchmarks are iffy, just try ones you want and see how well they perform the task you want.