AssemblyAI Universal 3.6 Pro vs Soniox STT-RT v5
AssemblyAI Universal 3.6 Pro is ahead on the Pipecat Dataset and time to first text, and Soniox STT-RT v5 on final-text delay. The Ocular Dataset is too close to call.
Last updated Methodology by Dileep Chagam
Accuracy vs speed
Each dot is one model. Bottom-left is best: fewer errors and a faster final transcript.
Which to pick
Pick AssemblyAI Universal 3.6 Pro for accuracy on clean, scripted audio (0.78 points lower) and showing words while the caller is still talking (710ms sooner). Pick Soniox STT-RT v5 for quick turn-taking once the caller stops (53ms sooner). For accuracy on real phone conversations, the gap is inside the margin of error, so neither is ahead.
The four metrics, side by side
- AssemblyAI Universal 3.6 Pro
- 1.77%1st
- Soniox STT-RT v5
- 2.54%8th
AssemblyAI Universal 3.6 Pro, 0.78 points lower
- AssemblyAI Universal 3.6 Pro
- 4.50%14th
- Soniox STT-RT v5
- 3.92%6th
Too close to call: inside the margin of error
- AssemblyAI Universal 3.6 Pro
- 485ms1st
- Soniox STT-RT v5
- 1.20s9th
AssemblyAI Universal 3.6 Pro, 710ms sooner
- AssemblyAI Universal 3.6 Pro
- 91ms3rd
- Soniox STT-RT v5
- 38ms1st
Soniox STT-RT v5, 53ms sooner
Lower is better on every metric. The small figure is each model's place in the whole field. What each metric measures.
- WER · Word error rate
- Share of words the model got wrong: replaced, missing or added. Lower is better.
- Pipecat Dataset
- 1,000 short public clips from Pipecat's STT benchmark dataset.
- Ocular Dataset
- 8 recordings from 4 real phone conversations, licensed from Ocular.
- TTFT · Time to first text
- How long after the audio starts streaming until the first words appear.
- TTFS · Final-text delay
- How long after the caller stops until the full transcript arrives. A voice agent waits for this before it replies.
First text and final text, at three percentiles
AssemblyAI Universal 3.6 Pro shows first text after 967ms and final text after 232ms, and Soniox STT-RT v5 shows first text after 1.71s and final text after 79ms, on a bad turn (P95) across the same streamed turns.
| Measure | Model | P50 | P90 | P95 |
|---|---|---|---|---|
| Time to first text | AssemblyAI Universal 3.6 Pro | 485ms | 966ms | 967ms |
| Soniox STT-RT v5 | 1.20s | 1.67s | 1.71s | |
| Final-text delay | AssemblyAI Universal 3.6 Pro | 91ms | 210ms | 232ms |
| Soniox STT-RT v5 | 38ms | 72ms | 79ms |
- P50 / P90 / P95 · Typical, slow and slowest turns
- Half of turns are faster than P50. Only 1 in 10 is slower than P90, and 1 in 20 slower than P95.
How each ranks among all 17 models
Place among the 17 models on each metric. Lower is better on all four.
| Metric | Universal 3.6 Pro | STT-RT v5 |
|---|---|---|
| Pipecat Dataset WER | 1.77%1st of 17 | 2.54%8th of 17 |
| Ocular Dataset WER | 4.50%14th of 17 | 3.92%6th of 17 |
| Time to first text | 485ms1st of 15 | 1.20s9th of 15 |
| Final-text delay | 91ms3rd of 17 | 38ms1st of 17 |
How much audio each model was tested on
| Model | Pipecat clips | Ocular recordings | Turns timed | Failed |
|---|---|---|---|---|
| AssemblyAI Universal 3.6 Pro | 1,000 | 8 | 206 | 0 |
| Soniox STT-RT v5 | 1,000 | 8 | 200 | 6 |
Only clips with usable text count, so the two totals can differ. How WER is scored.
Related comparisons
- AssemblyAI Universal 3.6 Pro vs AssemblyAI Universal 3.5 Pro
- AssemblyAI Universal 3.6 Pro vs Google Chirp 3
- AssemblyAI Universal 3.6 Pro vs Reson8
- AssemblyAI Universal 3.6 Pro vs GPT Realtime Whisper
- AssemblyAI Universal 3.6 Pro vs Cartesia Ink 2
- AssemblyAI Universal 3.6 Pro vs Speechmatics Linden
- AssemblyAI Universal 3.6 Pro vs Inworld STT-1
- AssemblyAI Universal 3.6 Pro vs GPT-4o Mini Transcribe
- AssemblyAI Universal 3.6 Pro vs Deepgram Nova-3
- AssemblyAI Universal 3.6 Pro vs Smallest Pulse
- AssemblyAI Universal 3.6 Pro vs GPT-4o Transcribe
- AssemblyAI Universal 3.6 Pro vs Deepgram Flux English
- AssemblyAI Universal 3.6 Pro vs Deepgram Flux Multilingual
- AssemblyAI Universal 3.6 Pro vs Gradium
- Soniox STT-RT v5 vs Speechmatics Linden
- Soniox STT-RT v5 vs Cartesia Ink 2
- Soniox STT-RT v5 vs Inworld STT-1
- Soniox STT-RT v5 vs GPT Realtime Whisper
- Soniox STT-RT v5 vs Reson8
- Soniox STT-RT v5 vs Google Chirp 3
- Soniox STT-RT v5 vs GPT-4o Mini Transcribe
- Soniox STT-RT v5 vs Deepgram Nova-3
- Soniox STT-RT v5 vs Smallest Pulse
- Soniox STT-RT v5 vs GPT-4o Transcribe
- Soniox STT-RT v5 vs Deepgram Flux English
- Soniox STT-RT v5 vs Deepgram Flux Multilingual
- Soniox STT-RT v5 vs Gradium
Frequently asked questions
Which is more accurate, AssemblyAI Universal 3.6 Pro or Soniox STT-RT v5?
AssemblyAI Universal 3.6 Pro is more accurate on the public clips. On the 1,000 public Pipecat clips, AssemblyAI Universal 3.6 Pro scores 1.77% and Soniox STT-RT v5 2.54%, so AssemblyAI Universal 3.6 Pro makes fewer errors by 0.78 points. On the eight licensed Ocular recordings, AssemblyAI Universal 3.6 Pro scores 4.50% and Soniox STT-RT v5 3.92%, with likely ranges of 3.66 to 5.34% and 2.60 to 4.86%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.
Is AssemblyAI Universal 3.6 Pro faster than Soniox STT-RT v5?
No. Soniox STT-RT v5 finishes the transcript 53ms sooner. AssemblyAI Universal 3.6 Pro has the final transcript 91ms after the speaker stops and Soniox STT-RT v5 38ms, at the median. Both are measured on the same 206 streamed turns.
Is AssemblyAI Universal 3.6 Pro or Soniox STT-RT v5 better for voice agents?
Soniox STT-RT v5 finishes the transcript 53ms 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.
Do AssemblyAI Universal 3.6 Pro and Soniox STT-RT v5 stream text while the speaker is talking?
Yes, both return partial text before the speaker stops. AssemblyAI Universal 3.6 Pro shows first text after 485ms and Soniox STT-RT v5 after 1.20s, at the median, so AssemblyAI Universal 3.6 Pro is 710ms sooner.