whatmodel.sh / stt / Chirp 3

Chirp 3

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

Chirp 3 is a closed-weights speech-to-text model from Google, released March 2025. Licence: proprietary.

60.8±1.7#26 of 162 speech-to-text modelsfrom 15 benchmark results across 10 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 Chirp 3, out of 176 with a score, then more from Google.

#modelscorewhereRAMreleased
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
33Gemini 3.5 Transcribe Live Google61±3api $0.0090/audio min2026-08
more from Google
11Gemini 3.5 Transcribe Google67±6api $0.0050/audio min2026-08
55Chirp 2 Google58±3api price unknown
110Chirp Google50±4comparison only2023-05

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

$ what it's good at

long-form transcription▁▁▄▁·▁▂█▄▁▁▁ #14 of 27 * · 10% of its score · 1 test

Trails Whisper large (v1), Whisper large-v2, Rev AI Machine (Reverb ASR) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Open ASR English long-form WER9.6% ✎▁▁▅▁·▃▅█▆▁▂▂ beats 14/30avg 12%▇▇▇▇ 10%typical
multilingual transcription▂··▃▄█▆▅▄▃·▂ #4 of 36 * · 4.4% of its score · 1 test

Beats Cohere Transcribe Arabic (07-2026), Cohere Transcribe (03-2026), Nova-3 here, though they rank higher overall. Trails SeamlessM4T v2 Large here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
mu-bench voice-agent caller utterances · batch UER (5 locales)11.9% ✎█·····▄····▄ beats 3/3avg 39%▇▇ 4.4%typical
other tests13 not in a task · 84% of its score
testresultvs the others in its rundifficultycountswhy
AfriSpeech-MultiBench · Parliament (legal domain) WER28.2% ✎▂▂····▂·▅·▃█ beats 3/12avg 52%▇▇▇▇▇▇▇ 17%sharp
Open ASR English short-form WER (March 2026)6.6% ✎▁▁▁▃··▁▁▂██▆ beats 67/86avg 8.2%▇▇▇▇ 10%blunt
RW-Voice-EQ ASR · 4 real-world English sets, mean WER8.7% ✎█▃▃▅▃···▃··▃ beats 6/9avg 15%▇▇▇ 8.3%typical
VoiceCodeBench · batch canonical token/entity match (CTEM)88.8% ✎▄···▄··▄▄▄·█ beats 3/6easy · avg 77%▇▇▇ 8.1%typical
GigaSpeechBench · 7 Arabic dialect WER mean38.2% ✎█··█·██··█·█ beats 5/5avg 79%▇▇ 6.1%typical
VoiceCodeBench · batch WER (300 English workplace segments)5.8% ✎▄·▄·▄·▄█···▄ beats 5/6avg 12%▇▇ 5.7%typical
GigaSpeechBench · 12 English domains entity B-WER mean14.7% ✎▄·······▄·█▄ beats 2/4avg 28%▇▇ 5.2%typical
Perle code-switching benchmark · 4-pair mean WER39.4% ✎█··········▄ beats 0/2avg 55%▇▇ 4.9%small run
5 more tests
testresultvs the others in its rundifficultycountswhy
BOTNOI report · FLEURS Thai CER10.3% ✎█··█·······█ beats 0/2avg 17%▇▇ 4.8%small run
GigaSpeechBench · 6 English accents WER mean17.2% ✎▄······▄··▄█ beats 2/4avg 36%▇▇ 4.3%typical
BOTNOI Thai telephony · CER13.4% ✎█·······█··█ beats 2/2avg 52%▇▇ 4.0%small run
koedesk FLEURS first-200 · 9-language headline error mean4.2% ✎▄·····▄··▄█▄ beats 2/5avg 7.9%▇ 3.5%typical
Quranic ASR Leaderboard · held-out recitation (600 clips) overall WER10.3% ✎▃·▅·▅▃▃·▃▃▃█ beats 11/12avg 24%▇ 1.9%blunt

$ where to run it

Through an API

api · price unknown · cloud.google.com ↗

$ how the score was worked out · Open ASR Leaderboard 21%, AfriSpeech-MultiBench (arXiv:… 17%, GigaSpeechBench: A Real-World… 16% and 7 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▇▇▇▇▇▇ 21%Apache-2.0
AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suite1▇▇▇▇▇ 17%CC BY 4.0
GigaSpeechBench: A Real-World Multilingual Speech-to-Text Benchmark3▇▇▇▇▇ 16%CC BY 4.0
VoiceCodeBench (arXiv:2608.28916v2) · structured-token recovery ASR benchmark2▇▇▇▇ 14%CC BY 4.0
BOTNOI Thai telephony ASR report (arXiv:2608.24916v1) · baseline CERs2▇▇▇ 9%CC BY 4.0
RW-Voice-EQ Bench (Hume AI)1▇▇ 8%CC BY 4.0
Benchmarking Commercial ASR Systems on Code-Switching Speech (arXiv:2605.19069v3)1▇ 5%CC BY 4.0
mu-bench: A Multilingual Utterance Transcription Benchmark (arXiv:2609.32082v1), Table 51▇ 4%CC BY 4.0
koedesk STT Bench v1 (Track A, FLEURS first-200)1▇ 4%CC BY 4.0
Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)1▇ 2%Apache-2.0
prior · pull toward the middle▇ 1%
weighted average= 60.8

$ 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