Granite Speech 3.3 8b
IBM · released Apr 14, 2025 · ibm-granite/granite-speech-3.3-8b
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 5
- Size
- 8.6B
Context measured in tokens
Our take
Written Sep 4, 2026Granite Speech 3.3 is an Apache-licensed speech-to-text model from IBM that turns audio into written words. It is built for speed and unusually strong on difficult audio such as meetings and accented speech, though its overall accuracy sits in the middle of the pack.
Pick this for fast transcription on capable hardware, or when your audio is messy — overlapping speakers and accented speech are both conditions where it beats most rivals. It is also the right choice if you need a permissive licence for research or commercial self-hosting. Skip it if you want a hosted API, if your audio is mostly podcasts or video, or if you need coverage beyond its five supported languages.
The case for it
- Processes audio at 264 times real time — faster than most measured models.
- Better than most on hard conditions: 7.7% word error on meetings against a field middle of 10.6%, and 9.1% on accented speech against 10.8%.
- Near-top accuracy on clean read-aloud audio at 1.1%, against a field middle of 1.4%.
- Apache 2.0 licence allows commercial use, modification and redistribution without restriction.
The case against it
- Middling overall accuracy at 5.3% average word error, between the field best of 4.4% and middle of 5.8%.
- Worse than most on podcasts and video at 8.6%, against a field middle of 8.2%.
- No commercial hosting offers in our data; you must build and run your own inference pipeline.
How good is it?
An open speech-to-text model for meeting and read-aloud recordings, though accented speech is harder for it.
- transcribing recordings of meetings in a roomRecorded meetings · 16th of 92
- transcribing clear recordings of people reading aloudClean read speech · 17th of 92
- transcribing speakers with a range of accentsAccented speech · 58th of 76
TranscriptionTurning speech into text3 of 5Open ASR WER · 41st of 76
94.7%
Misses roughly one word in 19, averaged over nine English test sets.
263×55th of 74
an hour of audio in 14 seconds, on the board's own hardware. Your machine will differ.
5
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.
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.
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. 15.7 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 23.7 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU) · 16 GB
Room to spare. 4.9 GB spare means a 10% error in the size would not change the answer.
These cards answer whether Granite Speech 3.3 8b 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 Granite Speech 3.3 8b
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
- ibm-granite/granite-speech-3.3-8b
- Architecture
- Dense
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- ibm-granite-granite-speech-3-3-8b