The Mercury 2.5 language model reportedly reaches 770 tokens per second, a speed well ahead of typical response rates. Tokens are the small chunks of text a model reads and writes, so a higher rate means text is generated and analysed much faster.
Speed changes what a model can be used for. Below a certain rate an AI feels like a slow correspondent; above it, interactions feel instant, and long documents or large batches can be processed without a noticeable wait.
What it means: fast models unlock uses that slow ones cannot, such as real-time assistance, live translation or processing thousands of records on a schedule. For automation, throughput often matters as much as intelligence, because a task that must finish overnight has no room for a slow model.