Universal 3 5 Pro
AssemblyAI
Speech to textTranscribes a recording into words
- Type
- Closed
- Input
- None held
- Output
- None held
- Cached
- None held
We don't hold a list price for this model yet · hosted only — no weights published
Our take
Written Sep 30, 2026Universal 3 5 Pro is a speech-to-text model with strong measured accuracy on clearly recorded English, and a real weak spot in meeting rooms. We list no download and no host for it, so there is no route to it we can point you to.
Reach for it on clearly recorded English speech — read-aloud audio, earnings calls, podcasts and video — where it sits among the better models we list, and on European-accented English in particular, where it places 5th of 92 on European-accented speech as of 28 Sep 2026. Skip it if your audio is meeting-room recordings with people talking over each other and a distant microphone, where it lands worse than most models we list.
The case for it
- Among the more accurate models we list on clean and lightly accented English: 1.1% of words wrong on read-aloud audiobooks, against a field middle of 1.5%, and 5th of 92 on European-accented speech as of 28 Sep 2026.
- Strong on corporate earnings calls, at 1.81% of words wrong with professional reference transcripts.
- Holds up on everyday internet audio: 7.62% of words wrong across podcasts, YouTube and audiobooks, better than most models on this condition.
The case against it
- Meeting-room audio is its weak condition: 10.63% of words wrong on real meetings with crosstalk and a distant microphone, worse than most models we list.
- No route to use it that we can point you to — we list no download for it and no host for it, so there is nothing here to run and nothing to call.
- Every accuracy figure we hold is English; 18 languages are listed but none of them is measured, so accuracy in the others is unverified in our data.
How good is it?
A speech-to-text model for turning recordings of meetings, podcasts and accented speakers into written text.
- turning spoken English into written textOpen ASR WER · 12th of 76
- transcribing speakers with a range of accentsAccented speech · 16th of 76
- transcribing podcasts and video audioPodcasts and video · 18th of 92
- transcribing clear recordings of people reading aloudClean read speech · 15th of 92
TranscriptionTurning speech into text4 of 5Open ASR WER · 12th of 76
95.7%
Misses roughly one word in 23, averaged over nine English test sets.
18
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 model8 scoresEvery figure we hold, from 8 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Where to get it
We hold no priced listing for Universal 3 5 Pro.
There is no copy to download and no host in our price data, so AssemblyAI is where to look. We watch OpenRouter, the provider APIs we track and the LiteLLM price set; this version appears in none of them, which is a gap in what we collect rather than a statement about what AssemblyAI sells.
Models people weigh against Universal 3 5 Pro
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
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- We don't hold a list price for this model yet — the gap is ours, not the lab's.
Licence and identifiers
What the licence allowsWe hold no licence record for this model. Inside are the identifiers you need to pull it — its Hugging Face repo where we have one, our slug and a machine-readable card.
Licence
We hold no licence row of this model's own. A source states its weights are not published, so the determination that governs it is the one for closed weights, API access only.
Identifiers
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- assemblyai-universal-3-5-pro