Models / IBM/ Granite Speech 3.3 8b

Granite Speech 3.3 8b

IBM · released Apr 14, 2025 · ibm-granite/granite-speech-3.3-8b

Speech to textTranscribes a recording into words

Input: audio. Output: text.InputOutput
Type
Open weightsApache License 2.0
Languages
5
Size
8.6B

Context measured in tokens

Our take

Written Aug 2, 2026

Granite Speech 3.3 is an Apache-licensed speech-to-text model from IBM that turns audio into written text. It is extremely fast on batch workloads and near-perfect on clean read-aloud English, though its accuracy widens sharply on harder audio and no hosted providers currently offer it.

Who should pick it

Pick this for high-throughput batch transcription of clean English audio where you need a permissive licence, or for regulated environments that require Apache-licensed self-hosted deployment. Skip it if you are transcribing meetings, podcasts, accented speech or video; if you need hosted inference; or if you need verified accuracy in French, German, Spanish or Portuguese.

The case for it

  • Extremely fast batch transcription at 264 times real time — an hour of audio in 14 seconds on the leaderboard's own hardware.
  • Truly permissive Apache 2.0 licence allows commercial use, modification and redistribution.
  • Near-perfect on clean read-aloud English, with about one word in ninety wrong.
  • Five languages supported: English, French, German, Spanish and Portuguese.

The case against it

  • Accuracy collapses on challenging audio: podcasts, video and accented speech are more than eight times worse than clean read speech, and recorded meetings are seven times worse.
  • No hosted offers available; you must self-host.
  • Every accuracy figure is English only; nothing we hold measures the other four languages.
00

How good is it?

TranscriptionTurning speech into text3.5 of 5Open ASR WER · 25th of 74

Words it gets right

94.7%

Misses roughly one word in 19, averaged over nine English test sets.

How fast it listens

264×46th of 62

an hour of audio in 14 seconds, on the board's own hardware. Your machine will differ.

Languages

5

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

Where it struggles
Read aloudaudiobooks, clean recording1.1%
Podcasts and videoeveryday internet audio8.6%
Accented speechspeakers from many countries9.1%
Meetingsa room, several people, far microphone7.7%

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

English · French · German · Spanish · Portuguese

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
Recorded meetings 11th of 74Clean read speech 15th of 74Accented speech 17th of 74Harder read speech 17th of 74Podcasts and video 48th of 74Financial calls 63rd of 74European-accented speech 63rd 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.
263.7independentsource ↗
5.3independentsource ↗
9.1independentsource ↗
3.5independentsource ↗
7.7independentsource ↗
8.6independentsource ↗
1.1independentsource ↗
2.5independentsource ↗
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_M5.4 / 24 GBest
Spare memory15.7 GB spare
Usable context66Kwhat the spare memory holds; no published limit on record
Decode speed155 tok/sest

Room to spare. 15.7 GB spare means a 10% error in the size would not change the answer.

One step upFits in memory

GeForce RTX 5090 · 32 GB

Weights at Q4_K_M5.4 / 32 GBest
Spare memory23.7 GB spare
Usable context131Kwhat the spare memory holds; no published limit on record
Decode speed275 tok/sest

Room to spare. 23.7 GB spare means a 10% error in the size would not change the answer.

On a MacFits in memory

Apple M1 (8-core GPU) · 16 GB

Weights at Q4_K_M5.4 / 16 GBest
Spare memory4.9 GB spare
Usable context16Kwhat the spare memory holds; no published limit on record
Decode speed9 tok/sest

Room to spare. 4.9 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
5.4 GBest
Fits in memory
Q5_K_M
6.4 GBest
Fits in memory
Q8_0
9.5 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

Models people weigh against Granite Speech 3.3 8b

03

When we formed this view

Dates behind this page

Aug 2, 2026BenchmarkScored 9.1 on Accented speechleaderboard
Aug 2, 2026BenchmarkScored 3.5 on Financial callsleaderboard
Aug 2, 2026BenchmarkScored 7.7 on Recorded meetingsleaderboard
Aug 2, 2026BenchmarkScored 4.5 on European-accented speechleaderboard
Aug 2, 2026BenchmarkScored 8.6 on Podcasts and videoleaderboard
Aug 2, 2026BenchmarkScored 1.1 on Clean read speechleaderboard
Aug 2, 2026BenchmarkScored 2.5 on Harder read speechleaderboard
Aug 1, 2026ListedListed on LLMapfirst indexed by our pipeline
Aug 1, 2026BenchmarkScored 263.7 on Open ASR RTFxleaderboard
Aug 1, 2026BenchmarkScored 5.3 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.
04

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

Architecture
Dense
Modality record
audio->text
Catalogue slug
ibm-granite-granite-speech-3-3-8b

Machine-readable model card (omc.json) →

Something wrong on this page? Tell us