There is no reason why this would run at “MP3 quality”, given that it would be a really bad idea to compress the audio data before running it through an algorithm. I would expect it’s at minimum CD quality, and perhaps better, depending on the fidelity of the A/D and D/A stages and the bandwidth of the algorithm.