AssemblyAI Universal 3.5 Pro vs Speechmatics Linden
AssemblyAI Universal 3.5 Pro is ahead on the Pipecat Dataset and time to first text, and Speechmatics Linden on final-text delay. The Ocular Dataset is too close to call.
AssemblyAI Universal 3.5 Pro has a newer release. See AssemblyAI Universal 3.6 Pro vs Speechmatics Linden.
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 3.5 Pro for accuracy on clean, scripted audio (0.53 points lower) and showing words while the caller is still talking (701ms sooner). Pick Speechmatics Linden for quick turn-taking once the caller stops (33ms 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 3.5 Pro
- 1.93%2nd
- Speechmatics Linden
- 2.46%7th
AssemblyAI Universal 3.5 Pro, 0.53 points lower
- AssemblyAI Universal 3.5 Pro
- 4.23%10th
- Speechmatics Linden
- 3.93%7th
Too close to call: inside the margin of error
- AssemblyAI Universal 3.5 Pro
- 489ms2nd
- Speechmatics Linden
- 1.19s8th
AssemblyAI Universal 3.5 Pro, 701ms sooner
- AssemblyAI Universal 3.5 Pro
- 180ms9th
- Speechmatics Linden
- 147ms8th
Speechmatics Linden, 33ms 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 3.5 Pro shows first text after 1.08s and final text after 692ms, and Speechmatics Linden shows first text after 2.52s and final text after 213ms, on a bad turn (P95) across the same streamed turns.
| Measure | Model | P50 | P90 | P95 |
|---|---|---|---|---|
| Time to first text | AssemblyAI Universal 3.5 Pro | 489ms | 966ms | 1.08s |
| Speechmatics Linden | 1.19s | 1.56s | 2.52s | |
| Final-text delay | AssemblyAI Universal 3.5 Pro | 180ms | 532ms | 692ms |
| Speechmatics Linden | 147ms | 192ms | 213ms |
- 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 3.5 Pro | Linden |
|---|---|---|
| Pipecat Dataset WER | 1.93%2nd of 17 | 2.46%7th of 17 |
| Ocular Dataset WER | 4.23%10th of 17 | 3.93%7th of 17 |
| Time to first text | 489ms2nd of 15 | 1.19s8th of 15 |
| Final-text delay | 180ms9th of 17 | 147ms8th of 17 |
How much audio each model was tested on
| Model | Pipecat clips | Ocular recordings | Turns timed | Failed |
|---|---|---|---|---|
| AssemblyAI Universal 3.5 Pro | 1,000 | 8 | 205 | 1 |
| Speechmatics Linden | 998 | 8 | 194 | 12 |
Only clips with usable text count, so the two totals can differ. How WER is scored.
Related comparisons
- AssemblyAI Universal 3.5 Pro vs AssemblyAI Universal 3.6 Pro
- Speechmatics Linden vs Cartesia Ink 2
- Speechmatics Linden vs Soniox STT-RT v5
- Speechmatics Linden vs Inworld STT-1
- Speechmatics Linden vs GPT Realtime Whisper
- Speechmatics Linden vs Reson8
- Speechmatics Linden vs Google Chirp 3
- Speechmatics Linden vs AssemblyAI Universal 3.6 Pro
- Speechmatics Linden vs GPT-4o Mini Transcribe
- Speechmatics Linden vs Deepgram Nova-3
- Speechmatics Linden vs Smallest Pulse
- Speechmatics Linden vs GPT-4o Transcribe
- Speechmatics Linden vs Deepgram Flux English
- Speechmatics Linden vs Deepgram Flux Multilingual
- Speechmatics Linden vs Gradium
Frequently asked questions
Which is more accurate, AssemblyAI Universal 3.5 Pro or Speechmatics Linden?
AssemblyAI Universal 3.5 Pro is more accurate on the public clips. On the 1,000 public Pipecat clips, AssemblyAI Universal 3.5 Pro scores 1.93% and Speechmatics Linden 2.46%, so AssemblyAI Universal 3.5 Pro makes fewer errors by 0.53 points. On the eight licensed Ocular recordings, AssemblyAI Universal 3.5 Pro scores 4.23% and Speechmatics Linden 3.93%, with likely ranges of 3.57 to 4.68% and 2.94 to 5.19%. 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 3.5 Pro faster than Speechmatics Linden?
No. Speechmatics Linden finishes the transcript 33ms sooner. AssemblyAI Universal 3.5 Pro has the final transcript 180ms after the speaker stops and Speechmatics Linden 147ms, at the median. Both are measured on the same 206 streamed turns.
Is AssemblyAI Universal 3.5 Pro or Speechmatics Linden better for voice agents?
Speechmatics Linden finishes the transcript 33ms 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 3.5 Pro and Speechmatics Linden stream text while the speaker is talking?
Yes, both return partial text before the speaker stops. AssemblyAI Universal 3.5 Pro shows first text after 489ms and Speechmatics Linden after 1.19s, at the median, so AssemblyAI Universal 3.5 Pro is 701ms sooner.