Qwen3 ASR 0.6B HF
Qwen · released Jun 26, 2026 · Qwen/Qwen3-ASR-0.6B-hf
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 30
- Size
- 0.8B
Context measured in tokens
Our take
Written Aug 2, 2026Qwen3 ASR is a tiny downloadable speech-to-text model with an Apache licence and support for 30 languages. It transcribes clean audio with very few errors and processes an hour of audio in roughly five seconds, though its accuracy falls sharply on difficult recordings and no commercial hosts currently offer it.
Pick this for lightweight self-hosted transcription where speed and licence permissiveness matter more than peak accuracy on messy audio. Use it when you need broad language coverage with English as your primary quality signal, or for very fast batch processing of clean recordings. Skip it if you need commercial hosting, if your audio is accented or from meetings, or if you need verified quality in non-English languages.
The case for it
- Processes one hour of audio in approximately five seconds on benchmark hardware, at 730.23 times real time.
- Strong on clean read-aloud audio, with 1.7% of words wrong.
- Apache 2.0 licence allows commercial use, modification and redistribution.
- 30 languages supported, including Chinese, English, Cantonese, Arabic, German, French, Spanish and Portuguese.
The case against it
- Accuracy collapses on challenging audio: 10.78% of words wrong on accented speech and 9.32% on recorded meetings, more than five times its clean-read rate.
- No commercial hosting available; must self-host.
- All accuracy figures we hold are English-only; the other 29 languages are unmeasured in our data.
How good is it?
TranscriptionTurning speech into text3 of 5Open ASR WER · 31st of 74
94.4%
Misses roughly one word in 18, averaged over nine English test sets.
730×32nd of 62
an hour of audio in 5 seconds, on the board's own hardware. Your machine will differ.
30
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, not to the board.
Which languages ↓ ↑
Chinese · English · Cantonese · Arabic · German · French · Spanish · Portuguese · Indonesian · Italian · Korean · Russian · Thai · Vietnamese · Japanese · Turkish · Hindi · Malay · Dutch · Swedish · Danish · Finnish · Polish · Czech · Filipino · Persian · Greek · Hungarian · Macedonian · Romanian
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.
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.
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
GeForce RTX 4090 · 24 GB
Room to spare. 20.9 GB spare means a 10% error in the size would not change the answer.
Apple M1 Pro (16-core GPU) · 32 GB
Room to spare. 22.1 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. 4.1 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
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 →
Models people weigh against Qwen3 ASR 0.6B HF
When we formed this view
Dates behind this page
Prices last checked 7h 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.
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
- Qwen/Qwen3-ASR-0.6B-hf
- Architecture
- Dense
- Modality record
- audio->text
- Catalogue slug
- qwen-qwen3-asr-0-6b-hf