Prompt speed

Also written: generation speed

Prompt speed is how many prompt tokens the model reads per second; generation speed is how many reply tokens it writes per second. Prompt reading is typically many times faster.