Yes. I've tried Sphinx implementations twice over the past few years. Google Cloud Speech and Nuance blew it out of the water.

It partially depends on dictionary size and the need for continuous recognition, but generally speaking there's a big divide between open source and paid services.