Gemma 4 E4B it
Google · released Mar 2, 2026 · google/gemma-4-E4B-it
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 140
- Size
- 8B
about 98K words of context
Our take
Written Sep 14, 2026Gemma 4 E4B it is Google's 7.94-billion-parameter speech-to-text model with an Apache licence and support for 140 languages. It is extremely fast and freely reusable, but its accuracy sits below the middle of the field on every English condition we measure.
Pick this when you need broad language coverage with a permissive licence, or when you must self-host and speed matters more than accuracy. Use it for offline batch transcription where 161 times real time on reference hardware lets you burn through large backlogs quickly. Skip it if you need top-tier accuracy on any audio type, or if you want a hosted provider to handle infrastructure.
The case for it
- Extremely fast: processes an hour of audio in 22 seconds on benchmark hardware.
- 140 languages supported, unusually broad for a downloadable transcription model.
- Apache 2.0 licence allows commercial use, fine-tuning and redistribution without restriction.
The case against it
- Below-median accuracy on every measured condition — more than double the middle rate even on clean read-aloud speech.
- Weak on everyday audio and accented speech, with error rates well above the field middle.
- No commercial hosting available; you must run it yourself.
How good is it?
An open speech-to-text model from Google that turns recordings into written text, though it trails most models on transcription.
- turning spoken English into written textOpen ASR WER · 70th of 76
- transcribing recordings of meetings in a roomRecorded meetings · 86th of 92
- transcribing speakers with a range of accentsAccented speech · 68th of 76
- transcribing podcasts and video audioPodcasts and video · 82nd of 92
- transcribing clear recordings of people reading aloudClean read speech · 86th of 92
TranscriptionTurning speech into text1 of 5Open ASR WER · 70th of 76
90.5%
Misses roughly one word in 11, averaged over nine English test sets.
161×61st of 74
an hour of audio in 22 seconds, on the board's own hardware. Your machine will differ.
140
Stated by the leaderboard; we do not hold the list itself.
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.
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. 16.2 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 24.2 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU) · 16 GB
Room to spare. 5.4 GB spare means a 10% error in the size would not change the answer.
These cards answer whether Gemma 4 E4B it 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.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked 7 days ago
- per 1M tokens
- $0.020 in / $0.10 out
- Context served
- 131K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| DeepInfrabf16Direct | $0.020 / $0.10checked 7 days ago | 131K | not measured | Unknown | Unknown | Unknown |
Across the 1 listings we hold: 0 say they do not train on prompts, 0 say they do and 1 does not say. 0 appear in the zero-retention registry we check; the rest are unknown to us.
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.
- 1 of 1 listings publishes no parameter list, so what its API accepts is unknown to us.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 1 listings does not say whether it trains on prompts.
- We hold no cached-input rate for any of its listings.
- We hold no batch or off-peak rate for any of its listings.
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
- google/gemma-4-E4B-it
- Architecture
- Dense
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- google-gemma-4-e4b-it