← Text-to-Speech

Qwen

qwen3-tts-flash

Measured 2026-07-03

Run qwen3-tts-flash on Speko →

One API, every model on this board.

Naturalness

1310

Win % · 95% CI

21%

16–27%

Drift

21

Synth p50

472ms

472–504 p50–p90

Robustness

0.15

Fined:

• fatal · currency — dropped the minus sign (3 of 5 transcripts)

→ we sent “The balance is -$12.50.”

→ it said “The balance is twelve point five dollars.”

• fatal · dates — never said “nineteen nineties” (31 of 35 transcripts)

→ we sent “This design dates to the 1990s.”

→ it said “This design dates to the one thousand nine hundred and nineties.”

• fatal · decimals — never said “zero point zero zero zero zero zero one” (24 of 31 transcripts)

→ we sent “The drift is 0.000001 per cycle.”

→ it said “The drift is oh point oh oh oh oh one per cycle.”

+ 12 more fined

Cost / 1M chars

$10.0

International (Singapore) list: $0.10 per 10,000 input characters, output audio not billed (alibabacloud.com/help/en/model-studio/model-pricing). A true per-character rate, no conversion. Mainland-China scope is ~$0.115/10k.

Share this result

Embed the live badge

Speko tts rank
Markdown
[![Speko tts rank](https://benchmarks.speko.ai/badge/tts/qwen-qwen3-tts-flash.svg)](https://benchmarks.speko.ai/tts/qwen-qwen3-tts-flash)
HTML
<a href="https://benchmarks.speko.ai/tts/qwen-qwen3-tts-flash"><img src="https://benchmarks.speko.ai/badge/tts/qwen-qwen3-tts-flash.svg" alt="Speko tts rank"></a>
URL
https://benchmarks.speko.ai/tts/qwen-qwen3-tts-flash