Audio
GA
Proprietary
Nova 3
Fast, accurate speech-to-text
Release2025-02
Context—
Max output—
ParametersUndisclosed
Local / VRAMAPI only
LicenseProprietary
Overview
Nova 3 is a model in the Audio category from Deepgram. It was released in 2025-02. The vendor is based in US. Best for: Fast, accurate speech-to-text.
Notes: Positioning from the vendor site; specs, access and pricing can change — verify before production use.
Strengths
- Low latency
- Speaker diarization
- Many languages
Limits
- Pay per minute
- Text output only
Specifications
| Vendor | Deepgram |
|---|---|
| Category | Audio |
| Release | 2025-02 |
| Status | GA |
| License | Proprietary |
| Commercial use | Not specified |
| Input modalities | Audio |
| Output modalities | Text |
| Prompt types | Text |
| Negative prompts | No |
| Max outputs | 1 |
| Local deployment | No |
| API access | Yes |
| Pricing | Usage-based API + free credit |
| Confidence | Medium |
| Vendor country | US |
| Vendor type | Media lab |
Pricing & hardware
Usage-based API + free credit
API-only: local GPU memory is not applicable. API access is available.