Dograh
Experiments  /  TTS verbalization benchmark  /  qwen3-tts
Model · EXP-001

qwen3-tts

Ranked 3 of 7 on the TTS verbalization benchmark. Strongest on stock ticker, weakest on hashtag or mention.

At a glance
Said right49%share of all 1,080 sentences
Rank3 of 7among the seven models
Cost per clip4.99scomputing time per audio clip
Likely range46%–52%where the true score most likely sits
Strengths and weak spots

Where qwen3-tts does best and worst

Its five strongest and five weakest types of text, by the share of sentences it said right.

Best at
Stock Ticker
0.92
Sports score
0.90
Decimal
0.82
Ordinal
0.82
Initialism or Acronym
0.80
Breaks on
Hashtag or Mention
0.00
URL or Email
0.16
Date
0.17
Address
0.17
Version Numbers
0.23
A real miss

Hashtag or Mention: what qwen3-tts actually said

The text
Follow #GreenBreak for updates during the conference.
Should be said as
Follow hashtag green break for updates during the conference.
qwen3-tts said
"follow groundbreak for updates during the conference"
All 27 types

Every type of text, best to worst

The share of sentences qwen3-tts said right on each of the 27 types. Each type has 40 test sentences.

Stock Ticker
0.92
Sports score
0.90
Decimal
0.82
Ordinal
0.82
Initialism or Acronym
0.80
Unit
0.72
Mathematical Expression
0.70
Geographic Coordinates
0.67
Currency
0.62
Chemical Formula
0.61
Abbreviation
0.59
License Plate or Serial Numbers
0.55
Vehicle or Product Code
0.51
Cardinal
0.50
Legal Reference
0.50
Phone Number
0.41
Fractions
0.40
Time
0.38
Biological Classification
0.33
Roman Numeral
0.31
ISBN
0.26
Musical Notation
0.23
Version Numbers
0.23
Address
0.17
Date
0.17
URL or Email
0.16
Hashtag or Mention
0.00