Murf Finch-Global vs GPT-4o Transcribe

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)
  • Finch-Global3.17% · 51ms
  • 4o Transcribe3.90% · 652ms

Murf Finch-Global is ahead of GPT-4o Transcribe on both datasets together, the Ocular Dataset and final-text delay. The Pipecat Dataset is too close to call.

Last updated Methodology by Dileep Chagam

Who is better at what

Finch-Global4o Transcribe
  • Overall WER3.33%6.85%
  • Pipecat Dataset WER3.17%3.90%
  • Ocular Dataset WER3.62%12.45%
  • Time to first text1.83sn/a
  • Final-text delay51ms652ms

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.

Finch-Global4o Transcribe
  • ELO rating998981–1017 · rank 5–21873834–934 · rank 22–23
  • Expected win rate67%33%
  • Ocular turns WER · 102 draws of 19482 won10 won
  • Pipecat WER · 745 draws of 999119 won135 won
  • Semantic WER · 161 draws of 19430 won3 won
  • Phone line · 101 draws of 19889 won8 won
  • Call-center noise · 68 draws of 198117 won13 won
  • Babble noise · 72 draws of 198104 won22 won
  • TV noise · 93 draws of 19864 won41 won
  • Reverb · 86 draws of 19781 won30 won
  • Short replies · 39 draws of 6928 won2 won
  • Entities · 40 draws of 483 won5 won

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

Finch-Global vs 4o Transcribe FAQ

  • Murf Finch-Global is more accurate on the public clips. On the 1,000 public Pipecat clips, Murf Finch-Global scores 3.33% and GPT-4o Transcribe 6.85%, so Murf Finch-Global makes fewer errors by 3.53 points. On the 300 licensed Ocular turns, Murf Finch-Global scores 3.17% and GPT-4o Transcribe 3.90%, with likely ranges of 2.84 to 3.51% and 3.43 to 4.42%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.

  • Final-text delay was not measured for one or both, so speed cannot be compared. Both are measured on the same 206 streamed turns.

  • Murf Finch-Global and GPT-4o Transcribe are level on the two figures a voice agent depends on most. A voice agent cannot reply until the final transcript arrives, and the Ocular recordings are real conversations.

  • Only Murf Finch-Global. GPT-4o Transcribe returns no text until the speaker stops, which matters if your agent reacts to partial transcripts.

Test your voice agent the same way.

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