AssemblyAI Universal 3.6 Pro vs GPT Realtime Whisper

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)
  • Universal 3.6 Pro1.77% · 91ms
  • Realtime Whisper2.16% · 543ms

AssemblyAI Universal 3.6 Pro is ahead on time to first text and final-text delay, and GPT Realtime Whisper on both datasets together. The Pipecat Dataset and The Ocular Dataset are too close to call.

Last updated Methodology by Dileep Chagam

Who is better at what

Universal 3.6 ProRealtime Whisper
  • Overall WER2.71%2.63%
  • Pipecat Dataset WER1.77%2.16%
  • Ocular Dataset WER4.50%3.53%
  • Time to first text485ms1.38s
  • Final-text delay91ms543ms

Lower is better on every measure. Too close to call means the likely ranges overlap. For costs, see our breakdown of AssemblyAI pricing.

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.

Universal 3.6 ProRealtime Whisper
  • ELO rating10211003–1031 · rank 2–1610091005–1020 · rank 5–16
  • Expected win rate52%48%
  • Ocular turns WER · 152 draws of 19626 won18 won
  • Pipecat WER · 889 draws of 99965 won45 won
  • Semantic WER · 187 draws of 1965 won4 won
  • Phone line · 149 draws of 19833 won16 won
  • Call-center noise · 119 draws of 19852 won27 won
  • Babble noise · 114 draws of 19855 won29 won
  • TV noise · 96 draws of 19813 won89 won
  • Reverb · 127 draws of 19857 won14 won
  • Short replies · 53 draws of 6812 won3 won
  • Entities · 46 draws of 502 won2 won

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

Universal 3.6 Pro vs Realtime Whisper FAQ

  • GPT Realtime Whisper is more accurate on the public clips. On the 1,000 public Pipecat clips, AssemblyAI Universal 3.6 Pro scores 2.71% and GPT Realtime Whisper 2.63%, so GPT Realtime Whisper makes fewer errors by 0.08 points. On the 300 licensed Ocular turns, AssemblyAI Universal 3.6 Pro scores 1.77% and GPT Realtime Whisper 2.16%, with likely ranges of 1.55 to 2.00% and 1.92 to 2.41%. 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 Universal 3.6 Pro finishes the transcript 894ms sooner. AssemblyAI Universal 3.6 Pro has the final transcript 485ms after the speaker stops and GPT Realtime Whisper 1.38s, at the median. Both are measured on the same 206 streamed turns.

  • AssemblyAI Universal 3.6 Pro finishes the transcript 894ms 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 3.6 Pro shows first text after 4.50% and GPT Realtime Whisper after 3.53%, at the median.

Test your voice agent the same way.

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