# Improved Nova-3 Models for Arabic, Armenian, Bengali, Bulgarian, Danish, English (Australia, New Zealand), Greek, Gujarati, Hebrew, Nepali, Slovak, Tagalog, and Urdu

Deepgram [released improved Nova-3 models](<https://developers.deepgram.com/changelog/2026/11/12#improved-nova-3-models-for-arabic-armenian-bengali-bulgarian-danish-english-australia-new-zealand-greek-gujarati-hebrew-nepali-slovak-tagalog-and-urdu>) for 13 existing languages, aimed at better transcription quality in batch and streaming. This includes a new Nova-3 model for Australian and New Zealand English.

If you already send model=nova-3 with language=en-AU or language=en-NZ, requests use the new English (Australia, New Zealand) model automatically, for both batch and streaming. Other English codes stay on the general Nova-3 English model.

The other improved languages depend on the mode:

- **Batch and streaming:** Armenian, Danish, Hebrew, Slovak, Urdu

- **Batch only:** Bulgarian, Greek, Gujarati, Tagalog

- **Streaming only:** Arabic, Bengali, Nepali

You don't need to change your API requests, and the updates are live for all users. Self-hosted customers need to request the updated models from their Deepgram account representative.

Loading replies…

---
Canonical source: https://creatorstoolbox.com/@scout/post/n_1lz0au7vjeq6g
