Skip to content
Agent Radar

OpenRouter adds a hosted Nemotron 3.5 Transcribe endpoint

Updated OpenRouter

Summary

The gateway lists a paid audio-transcription route, distinct from NVIDIA’s downloadable weights.

Tool
OpenRouter
Category
Agent infra
Date
Pricing
Not stated in the sources we cited.
Sources

OpenRouter now lists NVIDIA Nemotron 3.5 Transcribe as a hosted transcription endpoint, dated October 29. This is a gateway-availability update, not a second announcement of NVIDIA’s underlying model.

A hosted route for recorded audio

The model accepts audio and returns transcribed text. The catalog describes recordings up to one hour, support for 40 languages, speaker attribution, translation and prompt-guided vocabulary. Applications call the audio transcriptions endpoint using the model identifier nvidia/nemotron-3.5-transcribe.

Paid inference is separate from open weights

The displayed rate at review is $0.000033 per second. Downloadable weights do not make this hosted endpoint free, and the listing says EU in-region routing is not currently available. Check the current provider details before sending regional or sensitive workloads. We have not measured accuracy, response time or uptime.

See the OpenRouter hub.

More from OpenRouter

View all

Questions

What is OpenRouter?

According to its quickstart guide, OpenRouter exposes a chat completions endpoint that takes a model name and a list of messages. The guide also shows an @openrouter/sdk package, an @openrouter/agent package and a Python client.

Is OpenRouter open source?

Not stated in the cited sources. The quickstart excerpts do not give a license or say whether the service is open source.

How is OpenRouter used?

Its quickstart shows requests sent with an Authorization bearer API key to the chat completions endpoint using Python requests, fetch, curl, the OpenRouter SDKs, or the OpenAI SDK with the base URL pointed at OpenRouter.

Filed on Agent Radar. Claims in this note are limited to the sources above.