Models / SpeechBrain/ ASR Conformer Largescaleasr

ASR Conformer Largescaleasr

SpeechBrain · released Feb 6, 2025 · speechbrain/asr-conformer-largescaleasr

Speech to textTranscribes a recording into words

Input: audio. Output: text.InputOutput
Type
Open weightsApache License 2.0
Languages
1
Size
0.5B

Context measured in tokens

Our take

Written Aug 3, 2026

ASR Conformer Largescaleasr is a compact English-only speech-to-text model from SpeechBrain that processes audio extremely fast on benchmark hardware. Its Apache licence and small footprint suit self-hosted deployment, though accuracy falls sharply on messy real-world recordings and no commercial hosts currently offer it.

Who should pick it

Pick this for fast batch transcription of clean English audio where you can self-host — it processes an hour of audio in roughly 50 seconds and gets under 2% word error on read-aloud material. Use it for financial calls or European-accented speech at similar error rates. Skip it if your audio is from meetings, podcasts or heavily accented speakers, or if you need languages other than English.

The case for it

  • Extremely fast for its size class: processes an hour of audio in approximately 50 seconds on benchmark hardware.
  • Strong on clean, structured English audio, with under 2% word error on read-aloud material and under 4% on financial calls.
  • Fully open under Apache 2.0, with no cloud dependency required.
  • Compact at 0.5 billion parameters as listed.

The case against it

  • Accuracy collapses on challenging audio: error rate jumps to 15.57% on recorded meetings, more than eight times worse than on clean read speech.
  • English only; no measured data for any other language.
  • Zero commercial offers currently tracked, so self-hosting is the only path.
00

How good is it?

TranscriptionTurning speech into text1.5 of 5Open ASR WER · 65th of 74

Words it gets right

92.4%

Misses roughly one word in 13, averaged over nine English test sets.

How fast it listens

72.4×61st of 62

an hour of audio in 50 seconds, on the board's own hardware. Your machine will differ.

Languages

1

Listed on the model card. The accuracy above is English only.

Where it struggles
Read aloudaudiobooks, clean recording1.8%
Podcasts and videoeveryday internet audio9.7%
Accented speechspeakers from many countries14.2%
Meetingsa room, several people, far microphone15.6%

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.

Also scored, on boards we give no mark for
European-accented speech 45th of 74Harder read speech 53rd of 74Clean read speech 57th of 74Podcasts and video 63rd of 74Financial calls 66th of 74Accented speech 67th of 74Recorded meetings 67th of 74

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.
72.4independentsource ↗
7.6independentsource ↗
14.2independentsource ↗
3.7independentsource ↗
15.6independentsource ↗
9.7independentsource ↗
1.8independentsource ↗
4.1independentsource ↗
01

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

A card many people ownFits in memory

GeForce RTX 4090 · 24 GB

Weights at Q4_K_M0.3 / 24 GBest
Spare memory21.3 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed2771 tok/sest

Room to spare. 21.3 GB spare means a 10% error in the size would not change the answer.

One step upFits in memory

Apple M1 Pro (16-core GPU) · 32 GB

Weights at Q4_K_M0.3 / 32 GBest
Spare memory22.5 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed481 tok/sest

Room to spare. 22.5 GB spare means a 10% error in the size would not change the answer.

On a MacFits in memory

Apple M1 (8-core GPU, 8GB unified) · 8 GB

Weights at Q4_K_M0.3 / 8 GBest
Spare memory4.5 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed164 tok/sest

Room to spare. 4.5 GB spare means a 10% error in the size would not change the answer.

Your hardware
Checking your profile…

Memory use by level

Against a 24 GB card.

Q4_K_M
recommended
0.3 GBest
Fits in memory
Q5_K_M
0.4 GBest
Fits in memory
Q8_0
0.5 GBest
Fits in memory

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 →

02

When we formed this view

Dates behind this page

Aug 2, 2026BenchmarkScored 14.2 on Accented speechleaderboard
Aug 2, 2026BenchmarkScored 3.7 on Financial callsleaderboard
Aug 2, 2026BenchmarkScored 15.6 on Recorded meetingsleaderboard
Aug 2, 2026BenchmarkScored 4.1 on European-accented speechleaderboard
Aug 2, 2026BenchmarkScored 9.7 on Podcasts and videoleaderboard
Aug 2, 2026BenchmarkScored 1.8 on Clean read speechleaderboard
Aug 2, 2026BenchmarkScored 4.1 on Harder read speechleaderboard
Aug 1, 2026ListedListed on LLMapfirst indexed by our pipeline
Aug 1, 2026BenchmarkScored 72.4 on Open ASR RTFxleaderboard
Aug 1, 2026BenchmarkScored 7.6 on Open ASR WERleaderboard

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.
03

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

permissiveCommercial use allowed

Fully permissive: commercial use, redistribution, and derivatives allowed. Requires attribution and a copy of the license. Includes an express patent grant.

Identifiers

Modality record
audio->text
Catalogue slug
speechbrain-asr-conformer-largescaleasr

Machine-readable model card (omc.json) →

Something wrong on this page? Tell us