It's always useful to test how fast your models run on different backends -- JAX is often the fastest, but TensorFlow can surprise you!
There are often huge differences between backends. If you think that your go-to backend is "probably always fast enough", you're wrong