CekuraBench

Speechmatics Linden

Every modelSpeechmatics

Accuracy vs. speed

Lower and further left is better

Best quadrantBest picks
Final-text delay P50 (ms)
Pipecat Dataset WER (%)

Speechmatics Linden is 7th of 19 on the Ocular Dataset and 9th of 19 on final-text delay.

2.46% Pipecat WER (9th) · 3.93% Ocular WER (7th) · 1.19s First text (8th) · 147ms Final text (9th)Likely range, 95% confidence: Pipecat WER 2.19 to 2.73%, Ocular WER 2.94 to 5.19%Published price $0.30 per hour of streamed audio. Self-serve base rate, before the model-training and volume discounts. Source, checked Oct 8, 2026.

Last updated Methodology by Dileep Chagam

Speechmatics Linden against every other model

Lower is better on every measure. Hover a column for that model's full numbers.

  1. 2.9%
    Reson8
  2. 3.1%
    Ink 2
  3. 3.5%
    Pulse
  4. 3.5%
    Realtime Whisper
  5. 3.8%
    Chirp 2
  6. 3.9%
    STT-RT v5
  7. 3.9%
    Linden
  8. 4.2%
    Flux English
  9. 4.2%
    Chirp 3
  10. 4.2%
    Universal 3.5 Pro
  11. 4.4%
    Nova-3
  12. 4.5%
    Gradium
  13. 4.5%
    4o Mini Transcribe
  14. 4.5%
    Universal 3.6 Pro
  15. 4.5%
    Flux Multilingual
  16. 4.8%
    Scribe Pro
  17. 4.9%
    Scribe
  18. 10.6%
    STT-1
  19. 12.5%
    4o Transcribe
Glossary: WER, TTFT, TTFS, P50
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.
Likely range · 95% range
Where the model's true score probably falls, given how much audio was tested. Overlapping ranges are too close to call.
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.
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.

Where Speechmatics Linden struggled

On real calls, dropped words are its biggest source of error. On clean clips, wrong words lead.

55%

of its errors on the Ocular calls are words it left out.

Median across all 19 models: 53%

Mix of errors

Real calls (Ocular Dataset)

Clean clips (Pipecat Dataset)

TTFT and TTFS at P50, P90 and P95

Timed on the same 206 streamed Ocular turns as every other model. P90 and P95 show what a caller meets on a slow turn.

TTFS, final text

How long after the caller stops until the full transcript arrives. A voice agent waits for this before it replies.

20ms100ms500ms2s
Typical, P50
147ms
9th fastest of 19
Slow, P90
192ms
7th fastest of 19
Slowest, P95
213ms
7th fastest of 19

TTFT, first text

How long after the audio starts streaming until the first words appear.

200ms1s5s20s
Typical, P50
1.19s
8th fastest of 17
Slow, P90
1.56s
6th fastest of 17
Slowest, P95
2.52s
7th fastest of 17

Every other model's typical turn, on the same axis.

998 Pipecat clips · 8 Ocular recordings · 194 turns timed · 12 failed requests

Speechmatics Linden head to head

Every comparisonAll 19 modelsBenchmark source

Speechmatics Linden FAQ

  • Speechmatics Linden places highest on the Ocular Dataset, 7th of 19 at 3.93%. Its weakest placing is final-text delay, 9th of 19 at 147ms.

  • Speechmatics Linden transcribes the 1,000 public Pipecat clips at 2.46% word error rate and the eight licensed Ocular recordings at 3.93%. Word error rate counts incorrect, missing, and extra words against the reference transcript, so lower is better.

  • First text arrives after 1.19s at the median, and the transcript is final 147ms after the speaker stops. Both are measured on the same 206 streamed turns as every other model, and the page shows P90 and P95 as well, which is what a caller meets on a bad turn.

  • Every model runs the same two datasets under the same conditions: 1,000 public clips from Pipecat's STT benchmark dataset for accuracy at scale, and eight licensed conversation recordings from Ocular for real speech. Latency comes from 206 streamed turns, using each provider's own supported finalization contract. Speechmatics Linden returned usable text for 998 of the public clips.