Which models are available?

The models you can call depend on two things: your project’s region and language, and how each model runs. A model runs one of three ways, hosted by SLNG in your region, proxied through SLNG to the provider, or on your own provider account. In the dashboard these appear as the SLNG models and BYOK tabs on every model dropdown; in the API the route is part of the model ID.

Where a model can run

Prefer an SLNG-hosted model where one fits. It runs in your region and skips the hop to an outside provider, which is the lowest-latency path. Reach for a proxied or BYOK route when a model you need is not SLNG-hosted in your region, or when the provider account must stay in your name. The model ID names where the model runs:

Route	Runs on
`slng/deepgram/nova:3-en`	SLNG infrastructure, in your region
`deepgram/nova:3`	The provider, through SLNG. No key required
`deepgram/nova:3` + your credential	The provider, on your own account (BYOK)

The slng/ prefix is an API-only form. To run a model on your own account, register the key first; see Bring your own key.

See what’s available to you

Each model is offered per region and language, so the list you see is filtered to what your project supports.

Dashboard
API
CLI

Open your project and open any model dropdown on Listen, Think, or Speak. Each dropdown lists the models valid for the project’s region and language, split into the SLNG models and BYOK tabs.

A model dropdown open, showing the SLNG models and BYOK tabs — A model dropdown, with the SLNG models and BYOK tabs.

Call the catalog endpoint on your regional host to list the models available for that region. Only the host changes between regions; the path stays the same. To browse the same published data in the documentation, see All models.

curl "https://eu-west.api.slng.ai/v1/catalog/models?page_size=100" \
  -H "Authorization: Bearer $SLNG_API_KEY"

Response

{
  "items": [
    {
      "code": "deepgram/aura:2",
      "name": "Aura 2",
      "service_type": "TTS",
      "provider_code": "deepgram",
      "hosted_by": "Deepgram",
      "available_regions": ["au", "eu-west", "us-east"]
    }
  ],
  "meta": {
    "page": 1,
    "page_size": 100,
    "total": 14,
    "pages": 1
  }
}

From a coding agent, the voiceai CLI lists the currently deployed models. A ★ marks an SLNG-hosted model; the rest are proxied through SLNG.

voiceai models --stt

Output

STT models:
★ Nova 3 English (slng/deepgram/nova:3-en)
★ Nova 3 Spanish (slng/deepgram/nova:3-es)
★ Nova 3 Hindi (slng/deepgram/nova:3-hi)
★ Nova 3 Indonesian (slng/deepgram/nova:3-id)
★ Nova 3 Kannada (slng/deepgram/nova:3-kn)
★ Nova 3 Marathi (slng/deepgram/nova:3-mr)
★ Nova 3 Multi (slng/deepgram/nova:3-multi)
★ Nova 3 Telugu (slng/deepgram/nova:3-te)
★ Speechmatics Batch (slng/speechmatics/batch:15.0.0)
★ Speechmatics realtime v2 (slng/speechmatics/realtime:v2)
  Nova 3 (deepgram/nova:3)
  Nova 3 Medical (deepgram/nova:3-medical)
  Fish STT (fish/asr:default)
  Gradium STT (gradium/stt:default)
  Resonant 1 (reson8/reson8stt:v1)
  Saaras V3 (sarvam/saaras:v3)
  STT AI RT V3 (soniox/speech-ai:rt-v3)
  STT AI RT V4 (soniox/speech-ai:rt-v4)
  STT AI RT V5 (soniox/speech-ai:rt-v5)

Use --tts for text-to-speech models:

voiceai models --tts

Browse the full catalog

For the complete lists, see the model catalog: all models, speech to text, text to speech, LLM models, by region, and by language.

Voices are separate from text to speech models, and each model has its own set. Pick the model first, then a voice that belongs to it. For the voice catalog, see Cartesia voices and Gradium voices.

Next steps

Configure Listen, Think, and Speak
Bring your own key

Get started

Regions

Execution layer

Agent Builder

Models

Integrate

Where a model can run

See what’s available to you

Browse the full catalog

Next steps

​Where a model can run

​See what’s available to you

​Browse the full catalog

​Next steps

Where a model can run

See what’s available to you

Browse the full catalog

Next steps