Murf Finch-Global vs NVIDIA Nemotron 3.5 Transcribe

Compare

Accuracy and speed

Left gets fewer words wrong. Lower has the final text ready sooner. Nemotron 3.5 Transcribe doesn't support streaming, so it is not shown.

Accurate and fast0ms200ms400ms600ms800ms0%2%4%6%Words wrong on clean audio (Pipecat WER)
  • Finch-Global3.17% · 51ms
  • Nemotron 3.5 Transcribe1.89%

NVIDIA Nemotron 3.5 Transcribe is ahead of Murf Finch-Global on both datasets together and the Pipecat Dataset. The Ocular Dataset is too close to call.

Last updated Methodology by Dileep Chagam

Who is better at what

Finch-GlobalNemotron 3.5 Transcribe
  • Overall WER3.33%2.28%
  • Pipecat Dataset WER3.17%1.89%
  • Ocular Dataset WER3.62%3.03%
  • Time to first text1.83sn/a
  • Final-text delay51msn/a

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

Finch-Global vs Nemotron 3.5 Transcribe FAQ

  • NVIDIA Nemotron 3.5 Transcribe is more accurate on both datasets. On the 1,000 public Pipecat clips, Murf Finch-Global scores 3.33% and NVIDIA Nemotron 3.5 Transcribe 2.28%, so NVIDIA Nemotron 3.5 Transcribe makes fewer errors by 1.04 points. On the 300 licensed Ocular turns, Murf Finch-Global scores 3.17% and NVIDIA Nemotron 3.5 Transcribe 1.89%, so NVIDIA Nemotron 3.5 Transcribe makes fewer errors by 1.28 points. Word error rate counts incorrect, missing, and extra words, so lower is better.

  • They cannot be compared on speed. NVIDIA Nemotron 3.5 Transcribe was tested on whole recordings rather than a live stream, so its speed is not measured.

  • NVIDIA Nemotron 3.5 Transcribe makes fewer errors on real conversations. NVIDIA Nemotron 3.5 Transcribe was tested on whole recordings rather than a live stream, so its speed is not measured. A voice agent cannot reply until the final transcript arrives, and the Ocular recordings are real conversations.

  • NVIDIA Nemotron 3.5 Transcribe was tested on whole recordings rather than a live stream, so whether it streams partial text was not measured. Murf Finch-Global returns partial text before the speaker stops.

Test your voice agent the same way.

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