whatmodel.sh / stt / Parakeet Redux

Parakeet Redux

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

Parakeet Redux is an open-weights speech-to-text model from Moondream, released September 2026. Licence: cc-by-4.0.

Catalogue identity notes

moondream/parakeet-redux @2bf1286 (created 2026-09-18, CC BY 4.0): 'A 1.58-bit version of parakeet-tdt-0.6b-v3. Same architecture, same tokenizer, but every encoder weight is -1, 0 or +1. It fits in 178 MB'. Distinct from its full-precision sibling moondream/parakeet-ultra (post-trained, same 0.6B parameters) and from NVIDIA's parakeet-tdt-0-6b-v3. No parameter count is stated for the ternary release (HF safetensors metadata 148,923,399 counts packed U8 bytes, not weights), so params_b is left out.

Model information sources

Release date ↗

57.0±3.8#55 of 162 speech-to-text modelsfrom 3 benchmark results across 1 source

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 Parakeet Redux, out of 176 with a score, then more from Moondream.

#modelscorewhereRAMreleased
62Omnilingual ASR LLM 3B Meta57±2local~11GB2025-11
63Fun-ASR-MLT-Nano-2512 Alibaba (FunAudioLLM)57±4local~0.9GB2025-12
64Parakeet Redux Moondream57±4local~0.2GB2026-09
65Parakeet TDT 0.6B v3 NVIDIA57±1api $0.0015/audio min +local~2GB2025-08
66Kyutai STT 2.6B (en) Kyutai57±4local~6.7GB2025-06
more from Moondream
50Parakeet Ultra Moondream59±3local~0.7GB2026-09

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

$ what it's good at

other tests3 not in a task · 95% of its score
testresultvs the others in its rundifficultycountswhy
TED-LIUM long-form WER lab2.5% ✎█····█·█████ beats 5/6avg 4.6%▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 35%sharp · small run
FLEURS 25 European languages, mean WER (NVIDIA set) lab10.6% ✎█····█·····█ beats 1/2avg 18%▇▇▇▇▇▇▇▇▇▇▇▇ 31%small run
Granite 5.0 report · Open ASR public 7-set mean WER lab6.6% ✎█···█······█ beats 0/2avg 11%▇▇▇▇▇▇▇▇▇▇▇▇ 29%small run

$ where to run it

Other ways to run it

  1. Run it as published · needs 0.18 GB · from huggingface.co/moondream (official)
    pip install "moondream>=2.4.0" && python -c "import moondream as md; speech_cm = md.photon('moondream/parakeet-redux', device='cpu'); speech = speech_cm.__enter__(); print(speech.transcribe(audio='speech.wav')['text']); speech_cm.__exit__(None, None, None)"

Official weights and documentation: huggingface.co/moondream ↗

$ how the score was worked out · Moondream parakeet-ultra and… 95%

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
Moondream parakeet-ultra and parakeet-redux model cards (Benchmarks sections) · lab-reported3▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 95%CC BY 4.0
prior · pull toward the middle▇▇ 5%
weighted average= 57.0

$ 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