Deepgram Nova-3 Medical now transcribes ten languages in one request with language=multi (October 6, 2026)

What you get

Deepgram's changelog entry dated October 6, 2026 says the nova-3-medical model now supports multilingual transcription with language=multi. Speakers can switch between languages in the same medical recording.

The ten languages listed are Dutch (nl), English (en), French (fr), German (de), Hindi (hi), Italian (it), Japanese (ja), Portuguese (pt), Russian (ru) and Spanish (es).

It works for batch and streaming. It is available to hosted and self-hosted customers. Deepgram's models and languages page also lists multi under nova-3-medical with the same ten languages.

Prerequisites

A Deepgram API key in DEEPGRAM_API_KEY
An existing speech-to-text request to Deepgram, batch or streaming
Self-hosted only: the changelog says to request the updated models from your Deepgram account representative

Steps

In your transcription request, set the query parameters exactly as the changelog gives them: model=nova-3-medical&language=multi
Leave the rest of your request as it is. The changelog names no other required change.
If you run Deepgram self-hosted, ask your account representative for the updated models first.

The changelog has no full request or code sample, only the parameter string above.

Test it

Send a medical recording where the speaker switches between two of the listed languages, for example English and Spanish. Both languages should appear in the transcript.

Not confirmed

The pages do not state a price for this option.
They give no accuracy figures for the multilingual medical model.
They do not say which other features work together with language=multi on this model.
There is no full code sample because the announcement has none.

What you'll need

What you get

Prerequisites

Steps

Test it

Not confirmed