Hubert Large Ls960 ft
Meta · released Mar 2, 2022 · facebook/hubert-large-ls960-ft
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 1
- Size
- 0.3B
Context measured in tokens
Our take
Written Aug 3, 2026HuBERT Large Ls960 ft is a tiny, fast audio-to-text model from Meta that turns speech into written words. It is near-perfect on clean read-aloud audio and runs at over three thousand times real time, but its accuracy collapses on meetings and accented speech where roughly one word in three is wrong.
Pick this for batch transcription of pristine read-aloud audio where throughput matters more than perfection, or for research and education needing fully open, permissive weights. Use it to process large audio archives offline, where its speed enables massive parallelism on modest hardware. Skip it if you are transcribing meetings, accented speech, or any audio that is not studio-clean; or if you need a hosted service rather than self-hosting.
The case for it
- Extremely fast transcription throughput: over three thousand times real time, so an hour of audio processes in about 1.2 seconds on benchmark hardware.
- Near-perfect on pristine read-aloud audio, with a word error rate below 2%.
- Fully open under Apache License 2.0, and at 0.32 billion parameters small enough to run almost anywhere.
The case against it
- Collapses in real-world meeting audio, with a word error rate roughly eighteen times worse than on clean speech — about one word in three wrong.
- Struggles with accented speech at over seventeen times the clean-speech error rate, only marginally better than meetings.
- No commercial hosting available; you must self-host.
How good is it?
TranscriptionTurning speech into text1 of 5Open ASR WER · 73rd of 74
86.6%
Misses roughly one word in 7, averaged over nine English test sets.
3,024×14th of 62
an hour of audio in 1 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. 21.4 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. 22.6 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU, 8GB unified) · 8 GB
Room to spare. 4.6 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 →
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
- facebook/hubert-large-ls960-ft
- Architecture
- Dense
- Modality record
- audio->text
- Catalogue slug
- facebook-hubert-large-ls960-ft