ARK ASR 3B
AutoArk AI · released Jun 22, 2026 · AutoArk-AI/ARK-ASR-3B
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 19
- Size
- 4.1B
about 25K words of context
Our take
Written Sep 4, 2026ARK ASR 3B is a compact downloadable speech-to-text model from AutoArk AI that turns audio into written words. It runs at 484 times real time and scores better than most rivals on every English audio condition we measure, from clean read-aloud to noisy meetings.
Pick this for fast local transcription where speed matters — an hour of audio in about seven seconds on reference hardware. Use it for clean audiobook work, financial calls, or meeting transcription with a permissive Apache licence. Skip it if you need a hosted provider, verified accuracy beyond English, or the absolute lowest error rate on the overall average.
The case for it
- Faster than almost any model measured: 484 times real time on the benchmark harness.
- Better than most on every tested condition, from 1% errors on clean read-aloud to 7.9% on meetings.
- Apache 2.0 licence allows commercial use, modification and redistribution.
- Compact 4.1-billion-parameter weights with no mixture-of-experts complexity.
The case against it
- Not the absolute most accurate on the overall English average: 4.6% vs 4.4% for the best model.
- No hosted inference options currently listed; you must self-host.
- Accented speech and meetings rise to roughly 7–8 times the clean-audio error rate, though this gap is typical across the field.
How good is it?
An open speech-to-text model for turning recordings of meetings, podcasts and read-aloud speech into written text.
- turning spoken English into written textOpen ASR WER · 17th of 76
- transcribing recordings of meetings in a roomRecorded meetings · 20th of 92
- transcribing podcasts and video audioPodcasts and video · 4th of 92
- transcribing clear recordings of people reading aloudClean read speech · 9th of 92
TranscriptionTurning speech into text4 of 5Open ASR WER · 17th of 76
95.5%
Misses roughly one word in 22, averaged over nine English test sets.
481×49th of 74
an hour of audio in 7 seconds, on the board's own hardware. Your machine will differ.
19
Listed on the model card. The accuracy above is English only.
Percentage of words wrong on each set, lower better. Bars are scaled to this model's own worst case; the placing beneath each rate is against every model measured on that set.
Which languages ↓ ↑
Chinese · English · German · Japanese · French · Korean · Spanish · Polish · Italian · Romanian · Hungarian · Czech · Dutch · Finnish · Croatian · Slovak · Slovenian · Estonian · Lithuanian
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.
Each of these is the same transcription job on a different kind of recording, so together they say where it holds up and where it slips — not how closely it follows an instruction.
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.
Can you run it yourself?
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 18.9 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 26.9 GB spare means a 10% error in the size would not change the answer.
Apple M2 (8-core GPU, 8GB unified) · 8 GB
Room to spare. 2.1 GB spare means a 10% error in the size would not change the answer.
These cards answer whether ARK ASR 3B loads, not how fast it transcribes. Throughput figures for a transcription model come from its text decoder, so treat this as a fit answer rather than a speed one.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first. This is a fit answer: whether it loads, not how fast it transcribes.
Models people weigh against ARK ASR 3B
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
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.
Licence and identifiers
What the licence allowsApache License 2.0, 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
Apache License 2.0
Fully permissive: commercial use, redistribution, and derivatives allowed. Requires attribution and a copy of the license. Includes an express patent grant.
Identifiers
- Hugging Face
- AutoArk-AI/ARK-ASR-3B
- Architecture
- Dense
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- autoark-ai-ark-asr-3b