AssemblyAI STT 3.7 Realtime Agent vs Google Chirp 3

Compare

Accuracy and speed

Left gets fewer words wrong. Lower has the final text ready sooner.

Accurate and fast0ms200ms400ms600ms800ms0%2%4%6%Words wrong on clean audio (Pipecat WER)
  • STT 3.7 Realtime Agent1.68% · 47ms
  • Chirp 32.00% · 475ms

AssemblyAI STT 3.7 Realtime Agent is ahead of Google Chirp 3 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

Who is better at what

STT 3.7 Realtime AgentChirp 3
  • Overall WER2.41%2.76%
  • Pipecat Dataset WER1.68%2.00%
  • Ocular Dataset WER3.81%4.19%
  • Time to first text493ms4.48s
  • Final-text delay47ms475ms

Lower is better on every measure. Too close to call means the likely ranges overlap.

Clip by clip

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.

STT 3.7 Realtime AgentChirp 3
  • ELO rating10241001–1039 · rank 1–16979964–998 · rank 9–22
  • Expected win rate56%44%
  • Ocular turns WER · 160 draws of 19726 won11 won
  • Pipecat WER · 864 draws of 99983 won52 won
  • Semantic WER · 179 draws of 19714 won4 won
  • Phone line · 166 draws of 19823 won9 won
  • Call-center noise · 125 draws of 19745 won27 won
  • Babble noise · 86 draws of 19779 won32 won
  • TV noise · 114 draws of 19834 won50 won
  • Reverb · 143 draws of 19844 won11 won
  • Short replies · 59 draws of 7010 won1 won
  • Entities · 47 draws of 503 won0 won

Short replies are decided on the error count and entities on exact capture, with no draw margin. How the rating is worked out

STT 3.7 Realtime Agent vs Chirp 3 FAQ

  • AssemblyAI STT 3.7 Realtime Agent is more accurate on the public clips. On the 1,000 public Pipecat clips, AssemblyAI STT 3.7 Realtime Agent scores 2.41% and Google Chirp 3 2.76%, so AssemblyAI STT 3.7 Realtime Agent makes fewer errors by 0.34 points. On the 300 licensed Ocular turns, AssemblyAI STT 3.7 Realtime Agent scores 1.68% and Google Chirp 3 2.00%, with likely ranges of 1.47 to 1.90% and 1.77 to 2.25%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.

  • Yes. AssemblyAI STT 3.7 Realtime Agent finishes the transcript 3.99s sooner. AssemblyAI STT 3.7 Realtime Agent has the final transcript 493ms after the speaker stops and Google Chirp 3 4.48s, at the median. Both are measured on the same 206 streamed turns.

  • AssemblyAI STT 3.7 Realtime Agent finishes the transcript 3.99s 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 STT 3.7 Realtime Agent shows first text after 3.81% and Google Chirp 3 after 4.19%, at the median.

Test your voice agent the same way.

Cekura runs these simulated calls on your agent and scores every one.