Deepgram Nova Models by Deepgram
3 models2023–2025
Last refreshed 2026-10-17. Next refresh: weekly.
Capabilities
MultimodalAll models
Links
WebsiteAbout
Deepgram's Nova family contains general-purpose speech-to-text models optimized for high-accuracy transcription across batch and streaming workflows.
Decision facts
- Best fit
- audiospeech recognitionvision and multimodal work
- Capability starting point
- Deepgram Nova-3 with multimodal inputs
- Lowest tracked input
- Not tracked
- Closest related family
- AssemblyAI
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
2 in view1 retired
Deepgram Nova-3Current
Use when the workload needs speech recognition, multimodal inputs, and audio.
2025-02speech recognitionmultimodal inputsaudio
DeepgramCurrent
Use when the workload needs speech recognition, multimodal inputs, and audio.
2023-01speech recognitionmultimodal inputsaudio
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Deepgram Nova-3 | Use when the workload needs speech recognition, multimodal inputs, and audio. | 2025-02 | speech recognitionmultimodal inputsaudio | Current |
| Deepgram | Use when the workload needs speech recognition, multimodal inputs, and audio. | 2023-01 | speech recognitionmultimodal inputsaudio | Current |
Release Timeline
3 release groups2025-02
1 current
Deepgram Nova-3
Currentspeech recognitionmultimodal inputsaudio
2023-09
1 retired
Deepgram Nova-2
Replacedspeech recognitionmultimodal inputsaudio
2023-01
1 current
Deepgram
Currentspeech recognitionmultimodal inputsaudio
Replaced By
Replaced
Keep for legacy integrations; evaluate Deepgram Nova-3 before new work.
Specifications(3 models)
| Model | Released | Multimodal |
|---|---|---|
| Deepgram Nova-3 | 2025-02 | Yes |
| Deepgram | 2023-01 | Yes |