Nova 3
by Deepgram · Released 2025
Deepgram Nova 3 — production-grade STT with automatic punctuation and smart formatting. 11 languages with optional auto-detect.
Nova 3
Powered by Deepgram · Proprietary E2E ASR with diarization
Context Window
N/A
Parameters
Undisclosed
Max Output
N/A
Category
Speech to Text
Overview
Nova 3 is Deepgram's newest production STT model, offering high-accuracy transcription with rich post-processing. The CallMissed gateway calls it with automatic punctuation, smart formatting (numbers, dates, currencies, addresses), and language auto-detection.
Deployed on the CallMissed gateway, Nova 3 supports both batch (REST) and real-time (WebSocket) modes — this surface uses batch. For streaming voice agents, the WebSocket mode delivers interim results at sub-300ms latency.
At $0.50 per audio hour, it sits between Whisper (cheaper, English-leaning) and Saaras (Indian-languages-only). Pick Nova 3 when you need smart formatting or are not constrained to Indian languages. For diarization, domain-tuned models (medical, finance, meetings, and more), or 45+ language coverage, see the direct Deepgram line (deepgram-nova-3 and friends).
Pricing
| Metric | Price |
|---|---|
| Price /hour | ₹50.0000 |
1 credit = ₹1 = $0.01 USD. Prices shown from provider; CallMissed passes through with ~35% markup.
Key Highlights
- Smart formatting: numbers, dates, currencies
- Automatic punctuation and language auto-detect
- Batch + real-time WebSocket support
- Free on every plan
Benchmarks
| Benchmark | Score |
|---|---|
| Languages | 11 |
| Streaming latency | <300ms |
Technical Details
- Runs on the CallMissed gateway
- Batch: POST audio/mpeg → JSON transcript
- Streaming: WebSocket with interim_results, vad_events, utterance_end_ms
- Always-on: punctuate, smart_format; language auto-detected when not specified
Strengths
- Smart formatting saves post-processing pipelines
- Real-time WebSocket for sub-second voice agents
- Free on every plan
Limitations
- Only 11 languages vs Whisper's 99
- No diarization or domain-tuned modes on this route — use the direct deepgram-* line for those
- Streaming mode requires WebSocket plumbing (not exposed via REST)
- Higher per-hour cost than Whisper for general use
Use Cases
API Example
curl https://api.callmissed.com/v1/audio/transcriptions \ -H "Authorization: Bearer cm_YOUR_KEY" \ -F file=@call.mp3 \ -F model=nova-3 \ -F language=en-US
Endpoint: POST /v1/audio/transcriptions · Model ID: nova-3