Deepgram Nova-3 vs Soniox STT-RT v5

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)
  • Nova-33.46% · 103ms
  • STT-RT v52.54% · 38ms

Soniox STT-RT v5 is ahead on both datasets together, the Pipecat Dataset and final-text delay, and Deepgram Nova-3 on time to first text. The Ocular Dataset is too close to call.

Last updated Methodology by Dileep Chagam

Who is better at what

Nova-3STT-RT v5
  • Overall WER3.78%3.02%
  • Pipecat Dataset WER3.46%2.54%
  • Ocular Dataset WER4.39%3.92%
  • Time to first text1.06s1.20s
  • Final-text delay103ms38ms

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

Nova-3STT-RT v5
  • ELO rating971940–1015 · rank 5–221000986–1007 · rank 6–20
  • Expected win rate46%54%
  • Ocular turns WER · 163 draws of 19819 won16 won
  • Pipecat WER · 786 draws of 99975 won138 won
  • Semantic WER · 187 draws of 1987 won4 won
  • Phone line · 163 draws of 19818 won17 won
  • Call-center noise · 102 draws of 19735 won60 won
  • Babble noise · 83 draws of 19829 won86 won
  • TV noise · 133 draws of 19841 won24 won
  • Reverb · 149 draws of 19819 won30 won
  • Short replies · 61 draws of 705 won4 won
  • Entities · 42 draws of 505 won3 won

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

Nova-3 vs STT-RT v5 FAQ

  • Soniox STT-RT v5 is more accurate on both datasets. On the 1,000 public Pipecat clips, Deepgram Nova-3 scores 3.78% and Soniox STT-RT v5 3.02%, so Soniox STT-RT v5 makes fewer errors by 0.76 points. On the 300 licensed Ocular turns, Deepgram Nova-3 scores 3.46% and Soniox STT-RT v5 2.54%, so Soniox STT-RT v5 makes fewer errors by 0.92 points. Word error rate counts incorrect, missing, and extra words, so lower is better.

  • Yes. Deepgram Nova-3 finishes the transcript 131ms sooner. Deepgram Nova-3 has the final transcript 1.06s after the speaker stops and Soniox STT-RT v5 1.20s, at the median. Both are measured on the same 206 streamed turns.

  • It is a trade-off. Soniox STT-RT v5 makes fewer errors on real conversations, while Deepgram Nova-3 finishes the transcript 131ms sooner. 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. Deepgram Nova-3 shows first text after 4.39% and Soniox STT-RT v5 after 3.92%, at the median.

Test your voice agent the same way.

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