AssemblyAIAudio modelActive

Universal-3.6 Pro Realtime

PopularityRank 670 of 875

0out of 100same as last month

Summary

AssemblyAI describes Universal-3.6 Pro Realtime as a direct upgrade to Universal-3.5 Pro for voice agents and telephony. The model was trained for short utterances and difficult conditions such as loud rooms, accents, and phone lines. Its API automatically detects 32 languages and sends transcript text during the conversation.

AssemblyAI uses the identifier universal-3-6-pro at its existing streaming endpoint. The list price is $0.45 per hour of audio, with volume discounts.

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 223 published benchmark results

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

entity error rate (Show measurement, test conditions, and source)

Source value: 14.4
Score: 14.4
Metric: entity error rate
Unit: %
Benchmark version: source version not stated
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI announcement
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Same setup
Context: Source-specific observation; it does not establish a compatible cross-model cohort.

Source AssemblyAI announcement

exact transcript rate (Show measurement, test conditions, and source)

Source value: 87
Score: 87
Metric: score
Unit: %
Benchmark version: source version not stated
Category: audio
Direction: Higher is better
Source type: independent-evaluator
Evaluator: AssemblyAI announcement
Status: active
Retrieved at: 2026-10-04
Methodology: Independent evaluation. Same public run
Context: The source documents a separate evaluator setup; this observation is not merged with AssemblyAI's vendor benchmark.

Source AssemblyAI announcement

Coval pooled corpus-level STT word error rate (Show measurement, test conditions, and source)

Source value: 2.150440149608203
Score: 2.150440149608203
Metric: Pooled corpus-level speech-recognition word error rate, Coval pooled_value
Unit: %
Benchmark version: Coval rolling 30-day window, re-measured daily; route last measured 2026-10-16
Category: audio
Direction: Lower is better
Sample: 26332
Source type: official-leaderboard
Evaluator: Coval
Status: superseded
Retrieved at: 2026-10-16
Methodology: Official benchmark leaderboard. Coval endpoint key assemblyai/universal-3.6-pro on the exact /models/universal-3.6-pro route, hosted through AssemblyAI's own API. The directory's exact API model ID is universal-3-6-pro and its profile name is Universal-3.6 Pro Realtime. Coval's route identifies that same named AssemblyAI product, while its internal endpoint key uses different punctuation.
Context: This is Coval's pooled corpus-level STT WER, not the separate per-clip mean_value or a rounded leaderboard headline. Pooled WER and mean per-clip WER are different aggregations and should not be compared as the same score. The WER row's sample_count is used, not another metric's count or a roster total. The result describes Coval's hosted 30-day endpoint cohort, not local or native performance.

Source Coval

AssemblyAI English voice-agent benchmark, Universal 3.6 Pro code-switching WER mean (Show measurement, test conditions, and source)

Source value: 7.2
Score: 7.2
Metric: word error rate (WER), mean across five English to European language-pair directions
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. Exact API model ID universal-3-6-pro, public streaming API at default settings, in AssemblyAI's English voice-agent benchmark. The code-switching row is the mean across 20 public sets covering five English to European language pairs; the report lists EN to ES, DE, FR, IT, and PT pair results. The overall benchmark has 12,460 scripted scenarios, three noise environments, and 12 speaker groups, but a metric-specific scored N and uncertainty interval are not published.
Context: This is a vendor-reported benchmark. The 12,460 overall scenario plan and the 20 public sets are not substituted for the missing metric-specific scored denominator; no CI is published. The unversioned API route is identified, but the serving build is not.

Source AssemblyAI

AssemblyAI English voice-agent benchmark, Universal 3.6 Pro short-form WER (Show measurement, test conditions, and source)

Source value: 3.44
Score: 3.44
Metric: word error rate (WER) on English short-form speech
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. Exact API model ID universal-3-6-pro, public streaming API at default settings, in AssemblyAI's English voice-agent benchmark. English short-form WER is reported as a separate speech subset from overall English WER. The overall benchmark has 12,460 scripted scenarios, three noise environments, and 12 speaker groups, but the report gives no short-form scored N or uncertainty interval.
Context: This is a vendor-reported benchmark. The total 12,460 scenario count is not substituted for the missing short-form scored denominator; no CI is published. The unversioned API route is identified, but the serving build is not.

Source AssemblyAI

AssemblyAI English voice-agent benchmark, Universal 3.6 Pro short-answer semantic error (Show measurement, test conditions, and source)

Source value: 1.45
Score: 1.45
Metric: semantic wrong-answer rate on short user answers
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. Exact API model ID universal-3-6-pro in AssemblyAI's English voice-agent benchmark. The metric counts short-answer turns wrong by meaning: an equivalent answer such as yeah for yes is accepted, while a different or dropped answer is an error. The report does not state the scored subset size or uncertainty interval.
Context: This is a vendor-reported semantic subtask result, not WER and not a whole-benchmark score. The subset denominator and CI are not published. The unversioned API route is identified, but the serving build is not.

Source AssemblyAI

AssemblyAI English voice-agent no-speech false-transcript rate, Universal-3.6 Pro Realtime (Show measurement, test conditions, and source)

Source value: 0.27
Score: 0.27
Metric: Share of clips containing hold music, line noise or silence but no speech that returned transcript text; lower is better
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime release report, published 2026-09-29; 2,950 no-speech clips; model build not stated.
Category: speech-to-text / no-speech false activation
Direction: Lower is better
Sample: 2950
Source type: vendor-reported
Evaluator: AssemblyAI Universal-3.6 Pro Realtime release report
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. Exact product route universal-3-6-pro, production Universal-3.6 Pro Realtime endpoint. The report tested 2,950 clips of hold music, line noise and silence with no speech and counted whether any transcript text was emitted. It does not state request decoding controls or an immutable serving build for this result.
Context: Vendor-reported result from the model developer. It measures false text emission on no-speech audio, not transcription accuracy. No confidence interval or immutable endpoint build is stated; 2,950 is the reported clip denominator, not a number of independent speech utterances.

Source AssemblyAI Universal-3.6 Pro Realtime release report

Entity error rate: names (Show measurement, test conditions, and source)

Source value: 10.9
Score: 10.9
Metric: Entity error rate for names
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI Universal-3.6 Pro Realtime report
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. AssemblyAI's report says the evaluation used 12,460 scripted caller scenarios voiced by real speakers, three noise environments, and 12 speaker groups. Competitors were called through public streaming APIs at provider defaults with the same audio and scoring pipeline. This is a vendor-reported row; its row-specific denominator is not published.
Context: Vendor-reported, report-specific result. It is not an independently run cross-model cohort; exact API snapshot, per-row N, and confidence intervals are not disclosed unless stated separately.

Source AssemblyAI Universal-3.6 Pro Realtime report

Entity error rate: phone numbers (Show measurement, test conditions, and source)

Source value: 2.4
Score: 2.4
Metric: Entity error rate for phone numbers
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI Universal-3.6 Pro Realtime report
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. AssemblyAI's report says the evaluation used 12,460 scripted caller scenarios voiced by real speakers, three noise environments, and 12 speaker groups. Competitors were called through public streaming APIs at provider defaults with the same audio and scoring pipeline. This is a vendor-reported row; its row-specific denominator is not published.
Context: Vendor-reported, report-specific result. It is not an independently run cross-model cohort; exact API snapshot, per-row N, and confidence intervals are not disclosed unless stated separately.

Source AssemblyAI Universal-3.6 Pro Realtime report

Entity error rate: email addresses (Show measurement, test conditions, and source)

Source value: 17.2
Score: 17.2
Metric: Entity error rate for email addresses
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI Universal-3.6 Pro Realtime report
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. AssemblyAI's report says the evaluation used 12,460 scripted caller scenarios voiced by real speakers, three noise environments, and 12 speaker groups. Competitors were called through public streaming APIs at provider defaults with the same audio and scoring pipeline. This is a vendor-reported row; its row-specific denominator is not published.
Context: Vendor-reported, report-specific result. It is not an independently run cross-model cohort; exact API snapshot, per-row N, and confidence intervals are not disclosed unless stated separately.

Source AssemblyAI Universal-3.6 Pro Realtime report

Entity error rate: dates (Show measurement, test conditions, and source)

Source value: 18.6
Score: 18.6
Metric: Entity error rate for dates
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI Universal-3.6 Pro Realtime report
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. AssemblyAI's report says the evaluation used 12,460 scripted caller scenarios voiced by real speakers, three noise environments, and 12 speaker groups. Competitors were called through public streaming APIs at provider defaults with the same audio and scoring pipeline. This is a vendor-reported row; its row-specific denominator is not published.
Context: Vendor-reported, report-specific result. It is not an independently run cross-model cohort; exact API snapshot, per-row N, and confidence intervals are not disclosed unless stated separately.

Source AssemblyAI Universal-3.6 Pro Realtime report

Entity error rate: codes and IDs (Show measurement, test conditions, and source)

Source value: 10
Score: 10
Metric: Entity error rate for codes and IDs
Unit: %
Benchmark version: AssemblyAI Universal-3.6 Pro Realtime launch report, published 2026-09-29
Category: audio
Direction: Lower is better
Source type: vendor-reported
Evaluator: AssemblyAI Universal-3.6 Pro Realtime report
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. AssemblyAI's report says the evaluation used 12,460 scripted caller scenarios voiced by real speakers, three noise environments, and 12 speaker groups. Competitors were called through public streaming APIs at provider defaults with the same audio and scoring pipeline. This is a vendor-reported row; its row-specific denominator is not published.
Context: Vendor-reported, report-specific result. It is not an independently run cross-model cohort; exact API snapshot, per-row N, and confidence intervals are not disclosed unless stated separately.

Source AssemblyAI Universal-3.6 Pro Realtime report

Page 1 of 19

Popularity compared

How often the model is searched for, used and mentioned, compared with every model of its category.

Popularity index

0 to 100, higher is more popular

20 of 83 models

pyannote Community-173

Nemotron-3-Diarization57

Gemini 3.5 Transcribe52

Confucius4-R2T252

Granite Speech 5.0 470M TurboCTC52

Kokoro-82M51

Nova-351

Fish Audio S2 Pro51

NVIDIA Nemotron 3.5 Transcribe50

Granite Speech 5.0 470M TurboCTC NC50

Orukeet49

Lyria 3.547

Gemini 3.5 Live Translate46

Phonon-244

Aura-244

Gemini 3.1 Flash TTS44

MiniMax-Music341

Audio8-ASR-Infinite41

GPT-Realtime 2.141

Universal-3.6 Pro Realtime0

pyannote Community-1

Nemotron-3-Diarization

Gemini 3.5 Transcribe

Confucius4-R2T2

Granite Speech 5.0 470M TurboCTC

Kokoro-82M

Nova-3

Fish Audio S2 Pro

NVIDIA Nemotron 3.5 Transcribe

Granite Speech 5.0 470M TurboCTC NC

Orukeet

Lyria 3.5

Gemini 3.5 Live Translate

Universal-3.6 Pro Realtime

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 5, 2026

Universal-3.6 Pro RealtimeAmazon Nova 2 SonicEleven Multilingual STS v2Eleven Music v2.5Eleven Text to Sound v2

All specifications

Architecture

Model class: Streaming speech recognition; Source: Model class

Input and output

Input: Live audio; Source: Input
Output: Transcript; Source: Output
Languages: 32 languages with automatic detection; Source: Languages

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking: Unknown; no source
Watermarking technique: Not known; no source

Licence and access

Access: AssemblyAI API; Source: Access

API

API model ID: universal-3-6-pro; Source: API model ID

Further details

Streaming: Real-time model for AssemblyAI's existing streaming endpoint; Source: Streaming

Prices

Price: $0.45 per hour of audio; volume discounts availableSource

Sources and data date

Show all 27 sources

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

Without vendor claims and historical or disputed results; those stay labelled in the single values.

Type	Evidence and data date
Research status	Research date October 4, 2026. 21 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not published	metric-specific-sample-count-and-uncertainty Note AssemblyAI reports overall scenario coverage and subgroup definitions, but publishes no scored N or confidence interval for the code-switching mean, English short-form WER, or short-answer semantic error.
Unresolved	exact-serving-snapshot-and-unpublished-method-details Note The captured external source reports a benchmark-specific model or hosted route but does not establish an immutable currently served checkpoint. Unpublished row-level denominators and confidence intervals remain unknown; sample counts were entered only for cells that state them.
Additional source	AssemblyAI Universal-3.6 Pro Realtime (retrieved October 3, 2026; October 4, 2026; October 17, 2026; October 21, 2026) · AssemblyAI Universal-3.6 Pro Realtime · Editorial description reviewed October 3, 2026
Additional source	Coval benchmark (retrieved October 4, 2026)
Additional source	Coval (retrieved October 16, 2026; October 19, 2026; October 28, 2026; October 29, 2026; October 30, 2026; October 31, 2026; November 1, 2026; November 2, 2026; November 5, 2026; November 6, 2026)
Additional source	Pipecat STT benchmark (retrieved October 17, 2026)
Additional source	AssemblyAI (retrieved October 17, 2026; October 18, 2026)
Additional source	Omi Health Medical Speech-to-Text Benchmark (retrieved October 22, 2026)
Additional source	Model metadata (retrieved October 16, 2026)1 2
Additional source	Model metadata (retrieved October 17, 2026)1 2
Additional source	Model metadata (retrieved October 28, 2026)1 2 3 4 5
Additional source	Model metadata (retrieved November 5, 2026)1 2 3
Additional source	Model metadata (retrieved November 6, 2026)1 2 3
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Ollama: cumulative downloads for the popularity index
Additional source	Replicate: cumulative runs for the popularity index
Additional source	Google Trending Now: limited feed selection for Germany and the US
Additional source	Creative Commons Attribution 4.0