← Speech-to-Text 
Nari
Qwen3-ASR
Measured 2026-09-26
Run Qwen3-ASR on Speko →One API, every model on this board.
WER · batch
3.4%
2.1–4.9%
WER · streaming
Read speech
3.88%
Conversation
12.02%
Details
Read speech
n=1541 · FLEURS test+dev+train · vendor-direct
Conversation
10.0–14.4% · EdAcc n=283 · 60 speakers, 2 passes, 2026-10-01
Finalize
25ms
p50–p90 · US East · n=60 · vendor-direct · pre-close (30/30 flush-forced) · re-measured 2026-09-26 · ±22%
Commit-to-final: turn detection is off, so the probe commits at the measured speech end. Singapore reads 254ms.
Time to first token
1.75s
p50–p90 · US East · n=60 · vendor-direct · re-measured 2026-09-26 · ±0%
Cost / min
$0.001
Vendor list $0.06 per input audio hour (docs.narilabs.com/models-and-pricing, read 2026-09-26).
Share this result
Embed the live badge
Markdown
[](https://benchmarks.speko.ai/stt/nari-qwen3-asr)
HTML
<a href="https://benchmarks.speko.ai/stt/nari-qwen3-asr"><img src="https://benchmarks.speko.ai/badge/stt/nari-qwen3-asr.svg" alt="Speko stt rank"></a>
URL
https://benchmarks.speko.ai/stt/nari-qwen3-asr