Answer: In a contest that tested the new software against human translators, it came close to matching the fluency of humans for some languages.
Step-by-step explanation: Because they are close to matching fluency, they are very similar, and therefore comparable.