Models / AssemblyAI/ Universal 3 Pro

Universal 3 Pro

AssemblyAI

Speech to textTranscribes a recording into words

Input: audio. Output: text.InputOutput
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 4, 2026

Universal 3 Pro turns recorded speech into written text across 99 languages. It excels on clean read-aloud and European-accented formal audio, but struggles with meetings and heavily accented speech. No hosted providers or pricing are currently listed.

Who should pick it

Choose this for audiobooks, financial earnings calls, or European parliamentary-style speech where accuracy is strong. It also suits podcasts and internet video if a 7.5% word error rate is acceptable. Skip it if you are transcribing meetings, heavily accented international speech, or need a hosted API with known pricing.

The case for it

  • Among the most accurate on clean read-aloud audio at 1.1% word error rate.
  • Better than most on podcasts and video at 7.5% word error rate.
  • Covers 99 languages, far broader than most transcription models.
  • Strong on formal European-accented speech at 1.8% word error rate.

The case against it

  • Worse than most on heavily accented speech at 10.9% word error rate.
  • Worse than most on meeting audio at 11.1% word error rate.
  • No hosted inference options or pricing currently listed; speed is also unmeasured.
00

How good is it?

A speech-to-text model for turning podcasts, video audio and clear read-aloud recordings into written text.

Good at
  • transcribing podcasts and video audioPodcasts and video · 17th of 92
  • transcribing clear recordings of people reading aloudClean read speech · 15th of 92

EverydayGeneral questions and everyday reasoning

not measured

Not yet scored on Arena Text (overall).

CodingWriting and fixing code on its own

not measured

Not yet scored on Arena Coding.

AgenticPlanning, calling tools, staying on task

not measured

Not yet scored on Arena Agent.

WritingDrafting and rewriting prose

not measured

Not yet scored on Arena Creative Writing.

Other boards it appears on
Financial calls 3rd of 92European-accented speech 3rd of 92Harder read speech 14th of 92Clean read speech 15th of 92Podcasts and video 17th of 92Recorded meetings 54th of 92

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 model6 scoresEvery figure we hold, from 6 boards, with who ran it and a link to the source — including the boards no rating above is built on.
1.52source ↗
11.06source ↗
1.83source ↗
7.54source ↗
1.1source ↗
2.27source ↗
01

Where to get it

We hold no priced listing for Universal 3 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.

02

Models people weigh against Universal 3 Pro

03

When we formed this view

Recent changes

Aug 2, 2026BenchmarkScored 1.52 on Financial calls
What movedleaderboard
Aug 2, 2026BenchmarkScored 11.06 on Recorded meetings
What movedleaderboard
Aug 2, 2026BenchmarkScored 1.83 on European-accented speech
What movedleaderboard
Aug 2, 2026BenchmarkScored 7.54 on Podcasts and video
What movedleaderboard
Aug 2, 2026BenchmarkScored 1.1 on Clean read speech
What movedleaderboard
Aug 2, 2026BenchmarkScored 2.27 on Harder read speech
What movedleaderboard
Aug 1, 2026ListedListed on LLMap
What movedfirst indexed by our pipeline

Each 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.
04

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-pro

Machine-readable model card (omc.json) →

Something wrong on this page? Tell us