Higgs Audio v3 STT
Boson AI · released Mar 25, 2026 · bosonai/higgs-audio-v3-stt
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 1
- Size
- 2.7B
Context measured in tokens
Our take
Written Aug 2, 2026Higgs Audio v3 STT is a 2.68-billion-parameter English speech-to-text model from Boson AI with a permissive Apache licence. It excels at fast batch transcription of clean audio and financial calls, but accuracy drops sharply on conversational or accented speech and no hosted options exist.
Pick this for local English transcription where permissive licensing matters, especially financial calls at about one word in fifty wrong, or clean read-aloud at roughly one in a hundred. Use it when batch speed is critical: benchmark hardware processes an hour of audio in about half a minute. Skip it if you need multiple languages, hosted inference without self-hosting, or reliable accuracy on meetings, podcasts, accented speech, or video.
The case for it
- Best accuracy on financial calls among all conditions measured, at 2.07% word error rate.
- Extremely fast batch transcription: 110.12 times real-time on benchmark hardware.
- Truly permissive licence for local deployment: Apache 2.0 allows commercial use, modification and redistribution.
- Strong on clean prepared speech, with 1.07% error on clean read speech and 2.6% on harder read speech.
The case against it
- Accuracy collapses in conversational and accented scenarios: recorded meetings are 6.7 times worse than clean read speech, accented speech 7.7 times worse, and podcasts and video 7.0 times worse.
- English only, with no accuracy data for any other language and no hosted inference options in our catalogue.
- Inconsistent handling of non-native English: European-accented speech at 3.73% versus 8.26% on accented speech generally, a notable gap within the same category.
How good is it?
TranscriptionTurning speech into text4.5 of 5Open ASR WER · 5th of 74
95.4%
Misses roughly one word in 22, averaged over nine English test sets.
110×57th of 62
an hour of audio in 33 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, 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.
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. 19.6 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 27.6 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. 2.8 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 Higgs Audio v3 STT
When we formed this view
Dates behind this page
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.
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-stt
- Architecture
- Dense
- Modality record
- audio->text
- Catalogue slug
- bosonai-higgs-audio-v3-stt