Models / Useful Sensors/ Moonshine Streaming Tiny

Moonshine Streaming Tiny

Useful Sensors · usefulsensors/moonshine-streaming-tiny

Speech to textTranscribes a recording into words

Input: audio. Output: text.InputOutput
Type
Open weightsMIT License
Languages
1
Size
30M

Context measured in tokens

Our take

Written Aug 2, 2026

Moonshine Streaming Tiny is a 30-million-parameter English speech-to-text model that trades accuracy for extreme speed. It can process an hour of audio in under a second on benchmark hardware, but gets roughly one word in nine wrong overall.

Who should pick it

Pick this for offline or edge deployment where speed matters more than precision and English-only coverage is fine. Use it for projects needing a permissive open licence with minimal resource requirements, or for batch transcription of clean read-aloud audio on low-end hardware. Skip it if you need multiple languages, hosted availability, or reliable accuracy on meetings, podcasts, or accented speech.

The case for it

  • Extreme speed: processes one hour of audio in 0.82 seconds on benchmark hardware.
  • Strongest on clean read-aloud English, with a word error rate roughly 2.8 times better than its own average.
  • Truly permissive MIT licence allows commercial use, modification and redistribution without copyleft requirements.
  • Minimal footprint: 30 million parameters fit comfortably on edge devices.

The case against it

  • Accuracy collapses on challenging real-world audio, with word error rate jumping nearly fivefold on accented speech compared with clean read speech.
  • No hosted availability; users must self-host, and it supports English only.
00

How good is it?

TranscriptionTurning speech into text1 of 5Open ASR WER · 70th of 74

Words it gets right

88.8%

Misses roughly one word in 9, averaged over nine English test sets.

How fast it listens

4,375×8th of 62

an hour of audio in under a second, on the board's own hardware. Your machine will differ.

Languages

1

Listed on the model card. The accuracy above is English only.

Where it struggles
Read aloudaudiobooks, clean recording4%
Podcasts and videoeveryday internet audio12.5%
Accented speechspeakers from many countries19.5%
Meetingsa room, several people, far microphone16.9%

Percentage of words wrong on each set, lower better. Bars are scaled to this model's own worst case, not to the board.

The figures above come from the Open ASR Leaderboard, an independent public test that runs every model on the same recordings. It is the only measurement of transcription quality we know of, so there are no other scores to show.

Also scored, on boards we give no mark for
Podcasts and video 68th of 74Recorded meetings 69th of 74Accented speech 70th of 74Financial calls 70th of 74Clean read speech 72nd of 74European-accented speech 74th of 74Harder read speech 74th of 74

These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done, which is why they get no rating.

Every published score for this model9 scoresEvery figure we hold, from 9 boards, with who ran it and a link to the source — including the boards no rating above is built on.
4375.2independentsource ↗
11.2independentsource ↗
19.5independentsource ↗
5.7independentsource ↗
16.9independentsource ↗
12.5independentsource ↗
4independentsource ↗
11.5independentsource ↗
01

Can you run it yourself?

Fits in memory
weights load entirely on the card
Spills to system RAM
some weights offload; much slower
Too large
will not load even with offload
est
size is calculated; the verdict could change by 10%

Comfortable fit

A card many people ownFits in memory

GeForce RTX 4090 · 24 GB

Weights at Q4_K_M0 / 24 GBest
Spare memory21.6 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed44338 tok/sest

Room to spare. 21.6 GB spare means a 10% error in the size would not change the answer.

One step upFits in memory

Apple M1 Pro (16-core GPU) · 32 GB

Weights at Q4_K_M0 / 32 GBest
Spare memory22.8 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed7698 tok/sest

Room to spare. 22.8 GB spare means a 10% error in the size would not change the answer.

On a MacFits in memory

Apple M1 (8-core GPU, 8GB unified) · 8 GB

Weights at Q4_K_M0 / 8 GBest
Spare memory4.8 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed2617 tok/sest

Room to spare. 4.8 GB spare means a 10% error in the size would not change the answer.

Your hardware
Checking your profile…

Memory use by level

Against a 24 GB card.

Q4_K_M
recommended
0 GBest
Fits in memory
Q5_K_M
0 GBest
Fits in memory
Q8_0
0 GBest
Fits in memory

All 3 sizes here are calculated, not measured. We hold no measured file for this model, so each size comes from the parameter count and every verdict above inherits that uncertainty.

Check against your own machine → · All 71 devices, with every size →

02

When we formed this view

Dates behind this page

Aug 2, 2026BenchmarkScored 19.5 on Accented speechleaderboard
Aug 2, 2026BenchmarkScored 5.7 on Financial callsleaderboard
Aug 2, 2026BenchmarkScored 16.9 on Recorded meetingsleaderboard
Aug 2, 2026BenchmarkScored 8.1 on European-accented speechleaderboard
Aug 2, 2026BenchmarkScored 12.5 on Podcasts and videoleaderboard
Aug 2, 2026BenchmarkScored 4 on Clean read speechleaderboard
Aug 2, 2026BenchmarkScored 11.5 on Harder read speechleaderboard
Aug 1, 2026ListedListed on LLMapfirst indexed by our pipeline
Aug 1, 2026BenchmarkScored 4375.2 on Open ASR RTFxleaderboard
Aug 1, 2026BenchmarkScored 11.2 on Open ASR WERleaderboard

Prices last checked 6h ago

What we do not know about this model yet

  • We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
  • Nothing we hold says whether an endpoint streams, so we do not show it either way.
03

Licence and identifiers

What the licence allowsMIT License, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.

Licence

MIT License

permissiveCommercial use allowed

Fully permissive: do anything with attribution. No patent grant, unlike Apache-2.0.

Identifiers

Modality record
audio->text
Catalogue slug
usefulsensors-moonshine-streaming-tiny

Machine-readable model card (omc.json) →

Something wrong on this page? Tell us