Hopper
nemotron-asr
NVIDIA Nemotron 0.6B, served from Oregon
Measured 2026-10-08
Build voice agents on Speko →nemotron-asr is measured here for comparison — it is not one of the models Speko routes.
WER · batch
9.13%
95% CI 7.0–11.4 · n=50 · vendor-direct batch · FLEURS
The scorer counts spelled-out numbers as errors. On the 40 clips without numbers it reads 7.6%.
WER · streaming
Read speech
9.83%
Conversation
23.05%
Details
Read speech
n=1541 · FLEURS test+dev+train · vendor-direct
Scored with the same normalizer as batch, which counts spelled-out numbers as errors.
Conversation
20.0–26.1% · EdAcc n=283 · 60 speakers, 2 passes, 2026-10-08
6 of 566 readings came back empty, scored as all-deletion.
Finalize
656ms
p50–p90 · US East · n=200 · vendor-direct · re-measured 2026-10-08 · ±8%
Marker-limited: 68 of 200 readings land before the measured speech end, 28 by more than 0.5s (Modulate on the same clips: 27 and 12). Hopper serves from Oregon, so US readings include a cross-country round trip.
Time to first token
1.52s
p50–p90 · US East · n=200 · vendor-direct · re-measured 2026-10-08 · ±7%
Cost / min
$0.00083
Vendor list, $0.05 per audio hour.
Share this result
Embed the live badge
[](https://speko.ai/benchmarks/stt/hopper-nemotron-asr)
<a href="https://speko.ai/benchmarks/stt/hopper-nemotron-asr"><img src="https://speko.ai/benchmarks/badge/stt/hopper-nemotron-asr.svg" alt="Speko stt rank"></a>
https://speko.ai/benchmarks/stt/hopper-nemotron-asr