whatmodel.sh / stt / Scribe v1

Scribe v1

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Scribe v1 is a closed-weights speech-to-text model from ElevenLabs, released February 2025. Licence: proprietary.

Catalogue identity notes

ElevenLabs model_id 'scribe_v1' (launched 2025-02-25). Distinct from catalogue elevenlabs-scribe-v2. The ElevenLabs models page (observed 2026-09-28) lists scribe_v1 under deprecated models ('outclassed by v2'), so api is null.

60.9±2.4comparison only: no current run option verifiedfrom 7 benchmark results across 4 sources

Score within the Open ASR English short-form WER (March 2026) comparison set only.

$ how it compares

The speech-to-text models ranked either side of Scribe v1, out of 176 with a score, then more from ElevenLabs.

#modelscorewhereRAMreleased
28Canary-Qwen-2.5B NVIDIA61±2local~6.6GB2025-07
29Pulse (STT) Smallest.ai61±3api $0.0040/audio min
30Scribe v1 ElevenLabs61±2comparison only2025-02
31Chirp 3 Google61±2api price unknown2025-03
32Voxtral Small 24B (2507) Mistral61±2api $0.0040/audio min +local~59GB2025-07
more from ElevenLabs
12Scribe v2 ElevenLabs66±1api $0.0037/audio min2026-01
81Scribe v2 Realtime ElevenLabs55±3api $0.0065/audio min2025-11

see all speech-to-text models · find the ones that fit your machine

$ what it's good at

long-form transcription▁▁▅▁·▃▅█▆▁▂▃ #1 of 31 * · 20% of its score · 1 test

Beats Scribe v2, Cohere Transcribe (03-2026), Canary-Qwen-2.5B here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Open ASR English long-form WER5.9% ✎▁▁▅▁·▃▅█▆▁▂▃ beats 30/30avg 12%▇▇▇▇▇▇▇▇ 20%typical
other tests6 not in a task · 78% of its score
testresultvs the others in its rundifficultycountswhy
Bangla STT benchmark · BanSpeech 1,001-clip batch CER14.7% ✎█·▄··▄··▄·▄▄ beats 3/6avg 37%▇▇▇▇▇▇▇▇▇▇ 26%sharp
Open ASR English short-form WER (March 2026)6.9% ✎▁▁▁▃··▁▁▂██▆ beats 63/86avg 8.2%▇▇▇▇▇▇▇▇ 20%blunt
verli STT benchmark · LibriSpeech test-clean WER2.0% ✎▃·····▃▅█▃▅▅ beats 5/11avg 3.4%▇▇▇▇▇ 12%typical
Open ASR Leaderboard – Chilean Spanish · 3-dataset average WER7.3% ✎▂▂·····▂▂·█▃ beats 8/8avg 13%▇▇▇▇ 9.5%typical
verli STT benchmark · SPGISpeech 1,200-clip sample WER3.2% ✎▂·····▄▂█▂▂▄ beats 4/11avg 4.6%▇▇ 5.5%typical
verli STT benchmark · Earnings-22 WER (2,741 utt.)15.4% ✎▃·▃▅··▃▃▅▅█▃ beats 2/13avg 16%▇▇ 5.1%blunt

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · Open ASR Leaderboard 40%, Bangla Speech-to-Text Benchma… 26%, verliapp/stt-benchmark · inde… 22% and 1 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Open ASR Leaderboard2▇▇▇▇▇▇▇▇▇▇▇▇ 40%Apache-2.0
Bangla Speech-to-Text Benchmark v1.0 (Riverborn)1▇▇▇▇▇▇▇▇ 26%MIT
verliapp/stt-benchmark · independent STT accuracy benchmark3▇▇▇▇▇▇▇ 22%MIT
Open ASR Leaderboard – Chilean Spanish (idsudd / Alonso Astroza)1▇▇▇ 9%Apache-2.0
prior · pull toward the middle▇ 2%
weighted average= 60.9

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json