ClosedDeepgram/10 trustMetrics-checked

Deepgram Nova-2

Name: Deepgram Nova-2
Rating: 6

Deepgram's lean workhorse STT model: English-first batch and streaming transcription tuned for speed per dollar.

About

Nova-2 is Deepgram's second-generation speech-to-text model, released to early access in November 2023 and made generally available in mid-2024 with multilingual support, custom model training and on-premises deployment. Deepgram's provider-published figures claim a 30% average word error rate reduction over competitors, 5-40x faster prerecorded inference, and WER consistently below 10% across domains, with a median WER around 8.4% on real-world English sets (provider-measured, treat accordingly). It covers English plus six additional languages added in late 2024, with speaker diarization, smart formatting, word-level timestamps and filler-word detection. Nova-2 is still offered at unchanged rates, but Deepgram now steers new projects toward Nova-3.

Specs

Context window

Audio-duration based; streaming and batch, no token context

Pricing

$0.35/hour streaming pay-as-you-go; batch from $0.0043/min (provider list)

Free quota

$200 free credit for new accounts

API access

Deepgram Speech-to-Text API (model=nova-2) for streaming and prerecorded audio; self-hosted/on-prem containers available

Weights

Closed — API only

Released

Not disclosed

Benchmarks

Provider-reported: 30% average WER reduction vs competitors; median WER ~8.4% on real-world English sets.

Independent: no third-party reproduction found — treat all figures as provider-claimed.

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

call-center-analyticsvoice-agentsmeeting-transcriptionbatch-audio-transcription

Trust score — 6/10

/10 trustMetrics-checked

Scored 6/10 (metrics tier, metrics-v1). Evidence: official Deepgram launch and Nova-2-vs-Nova-3 comparison pages verified live; current pricing page confirms Nova-2 still offered at unchanged rates. Strengths: wide production adoption (60M+ audio minutes in the first month per Deepgram), full feature set, on-prem option. Docked for: (1) proprietary with no weights; (2) every accuracy and speed figure is provider-measured with no independent reproduction; (3) superseded as the recommended default by Nova-3. License: proprietary.

How the trust score is built

Full methodology →

Adoption · stars/downloads, log-scaled vs peers: 8.0/10
Maintenance · recency, releases, issue load: 7.0/10
Openness · license + docs: 1.0/10
Manual review · sanity read + red-flag check: 8.0/10

· Early access Nov 2023; GA mid-2024 with multilingual + custom training + on-prem
· Provider claims 30% avg WER reduction vs competitors, sub-10% WER across domains
· Still offered ($0.35/hr streaming PAYG) but Nova-3 is the recommended default
· $200 free credit for new accounts