Nemotron 3.5 Transcribe via the Tokenator API

What the model is for

Nemotron 3.5 Transcribe is an open-weight speech recognition model from NVIDIA, a 30B mixture-of-experts with about 4B active parameters. It transcribes recordings of up to one hour in 40 languages, with speaker attribution, speech translation, and prompt-guided vocabulary and style.

Specification

Full nameNemotron 3.5 Transcribe
API IDnemotron-3.5-transcribe
Model vendorNVIDIA
TypeSpeech to text
Context window1M tokens
Response formatsjson
Default formatjson
Base price10 tokens per second of input audio
Tokenator usage multiplier1×
BadgeNEW
Input modalitiesaudio
Supported API formats/v1/audio/transcriptions
Upstream providers1
Current statusavailable
Data updated2026-11-06

Upstream providers

ProviderMultiplier
Unified LLM API1×

A request goes to the first available provider by priority; if it fails, Tokenator switches to the next one.

Example request

curl
curl https://api.tokenator.top/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-your-tokenator-key" \
  -F 'model=nemotron-3.5-transcribe' \
  -F 'file=@recording.wav' \
  -F 'response_format=json'

FAQ about Nemotron 3.5 Transcribe

What is the API ID of Nemotron 3.5 Transcribe?

nemotron-3.5-transcribe — put this into the model field of your request.

How do I call this model?

Send POST /v1/audio/transcriptions with the model's API ID. Request examples and parameters are in the audio documentation.

Why is the model absent from /v1/models?

Audio models have a separate catalog at GET /v1/audio/models. Use ?task=tts for synthesis or ?task=stt for transcription; use ?task=music for music. Chat endpoints refuse audio models.

Which API key should I use?

Use your regular Tokenator key with access to this model. Send it in the Authorization: Bearer sk-your-tokenator-key header.

How do I upload an audio file?

Send multipart/form-data with the API ID in model and one non-empty audio file up to 25 MiB in file. Let the client set Content-Type with the boundary.

How do I get a plain-text transcript?

Set response_format=text. The default response is JSON with a text field; verbose_json includes extra details when the model returns them.

Can I specify the recording language?

Send language when the model supports it. Support for prompt and other parameters also depends on the model; see the audio documentation.