Left gets fewer words wrong. Lower has the final text ready sooner.
AssemblyAI Universal 3.6 Pro is ahead of AssemblyAI Universal 3.5 Pro on both datasets together, time to first text and final-text delay. The Pipecat Dataset and The Ocular Dataset are too close to call.
Last updated Methodology by Dileep Chagam
Lower is better on every measure. Too close to call means the likely ranges overlap. For costs, see our breakdown of AssemblyAI pricing.
On every clip both models were scored on, the lower WER wins, and WERs within 5 points of each other are a draw; in the rating, a wider win counts for more. Expected win rate is what the two ELO ratings predict for a match.
Short replies are decided on the error count and entities on exact capture, with no draw margin. How the rating is worked out
Same clips, same turns, both releases.
Time from the speaker stopping to the final transcript.
19 turns in 20 are final within this.
Out of the same 205 turns.
On the same 1,000 clips.
19 turns in 20 show first words within this.
AssemblyAI Universal 3.6 Pro is more accurate on the public clips. On the 1,000 public Pipecat clips, AssemblyAI Universal 3.5 Pro scores 2.72% and AssemblyAI Universal 3.6 Pro 2.71%, so AssemblyAI Universal 3.6 Pro makes fewer errors by 0.01 points. On the 300 licensed Ocular turns, AssemblyAI Universal 3.5 Pro scores 1.93% and AssemblyAI Universal 3.6 Pro 1.77%, with likely ranges of 1.69 to 2.18% and 1.55 to 2.00%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.
AssemblyAI Universal 3.6 Pro is more accurate on the public clips. On the 1,000 public Pipecat clips, AssemblyAI Universal 3.5 Pro scores 2.72% and AssemblyAI Universal 3.6 Pro 2.71%, so AssemblyAI Universal 3.6 Pro makes fewer errors by 0.01 points. On the 300 licensed Ocular turns, AssemblyAI Universal 3.5 Pro scores 1.93% and AssemblyAI Universal 3.6 Pro 1.77%, with likely ranges of 1.69 to 2.18% and 1.55 to 2.00%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.
No. AssemblyAI Universal 3.6 Pro finishes the transcript 4ms sooner. AssemblyAI Universal 3.5 Pro has the final transcript 489ms after the speaker stops and AssemblyAI Universal 3.6 Pro 485ms, at the median. Both are measured on the same 206 streamed turns.
AssemblyAI Universal 3.6 Pro finishes the transcript 4ms sooner, and the two are too close to call on real conversations. A voice agent cannot reply until the final transcript arrives, and the Ocular recordings are real conversations.
Yes, both return partial text before the speaker stops. AssemblyAI Universal 3.5 Pro shows first text after 4.23% and AssemblyAI Universal 3.6 Pro after 4.50%, at the median.
Cekura runs these simulated calls on your agent and scores every one.