Higgs Audio v3 8b STT v2
Boson AI · released Apr 27, 2026 · bosonai/higgs-audio-v3-8b-stt-v2
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 1
- Size
- 8.9B
Context measured in tokens
Our take
Written Sep 17, 2026Higgs Audio v3 8b STT v2 is a speech-to-text model you can download and run yourself, and it is accurate on clean and everyday audio. Its weak spot is accented speech, and every accuracy figure we hold is English only.
Use it for clean read-aloud audio, podcasts, video and everyday internet recordings, where it sits better than most models on those conditions, and for clearing a long backlog — an hour of audio takes about 26 seconds on the leaderboard's own hardware. The licence allows commercial use, changes and redistribution (Apache License 2.0). Skip it if your audio is mostly accented speech, or if you need a language other than English.
The case for it
- 1% of words wrong on clean read-aloud recordings, better than most models on that condition, where the field's best is 0.9% and the middle 1.5%.
- 7.1% of words wrong on podcast and video recordings, better than most models on that condition, where the field's best is 6.1% and the middle 8.3%.
- 8.4% of words wrong on real meeting recordings, better than most models on that condition, where the field's best is 6.1% and the middle 10.4%.
- 137 times real time on the leaderboard's own hardware, so an hour of audio takes about 26 seconds there — your machine may differ.
The case against it
- 11.7% of words wrong on accented speech, worse than most models on that condition, where the field's best is 4.1% and the middle is 7.9%, so a meeting with accented speakers is a qualified use.
- English only: all accuracy figures we hold are English, so nothing here measures another language.
- We list no host for it, so running it yourself is the only route we can point you to.
How good is it?
An open speech-to-text model for turning podcasts, video and clear read-aloud recordings into text, though accents are its weak spot.
- transcribing podcasts and video audioPodcasts and video · 5th of 92
- transcribing clear recordings of people reading aloudClean read speech · 5th of 92
- transcribing speakers with a range of accentsAccented speech · 65th of 76
TranscriptionTurning speech into text3.5 of 5Open ASR WER · 35th of 76
95%
Misses roughly one word in 20, averaged over nine English test sets.
137×65th of 74
an hour of audio in 26 seconds, on the board's own hardware. Your machine will differ.
1
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.
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. 15.5 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 23.5 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU) · 16 GB
Room to spare. 4.7 GB spare means a 10% error in the size would not change the answer.
These cards answer whether Higgs Audio v3 8b STT v2 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 Higgs Audio v3 8b STT v2
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
- bosonai/higgs-audio-v3-8b-stt-v2
- Architecture
- Dense
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- bosonai-higgs-audio-v3-8b-stt-v2