AssemblyAI Universal 3.5 Pro shows 1.93% word error rate in a 15-model test

Tarush Agarwal reported that a speech-to-text benchmark ran 15 models through the same audio and pipeline. On 1,000 read-aloud clips from Pipecat FLEURS, AssemblyAI Universal 3.5 Pro had the lowest word error rate at 1.93% and the fastest median time to first text at 489 ms. The measurements are in the speech-to-text benchmark results.