CheckAPICost 2026 INDEX

Deepgram • Speech to Text (STT)

Deepgram Nova 3 API Pricing & Rate Card

Q: What is the token pricing for Deepgram Nova 3 in 2026?

Deepgram Nova 3 is billed at $0.0043/min per million input tokens and $0.00 per million output tokens under standard API tiers.

Q: What is the context window capacity of Deepgram Nova 3?

Deepgram Nova 3 supports up to Streaming tokens, allowing long-document ingestion and extensive conversational history.

Q: Does Deepgram Nova 3 support prompt caching or batch discounts?

Yes, prompt caching reduces recurring prefix token costs to roughly $0.00 per million tokens for cached blocks.

Complete 2026 developer pricing guide for Deepgram Nova 3. Benchmark token economics, context window limits, prompt caching discounts, and live inference latency across official provider endpoints.

⚡ 60-Second Speech-to-Text Verdict ✓ Verified October 2026

Automated Speech Recognition & Diarization Unit Economics

Speech Recognition Decision Verdict: Deepgram Nova-3 delivers the industry's lowest word-error rate (WER) and cost-per-minute for speech recognition: Pre-recorded audio costs $0.0043/minute ($0.26/audio hour) on Pay-as-you-go, while real-time streaming audio costs $0.0059/minute ($0.35/audio hour) with smart formatting and diarization included. Nova-3 is 40% cheaper than OpenAI Whisper API ($0.006/min) while streaming at sub-300ms latency. Compare speech tools in our Deepgram vs Whisper Showdown and Meeting Notetaker Bot Stack.

Pre-Recorded Audio $0.0043 / Min ($0.26 / Hr)

Real-Time Streaming $0.0059 / Min ($0.35 / Hr)

Whisper API Comparison 40% Cheaper than Whisper

Real-Time Latency <300ms Streaming ASR

Input Token Rate

$0.0043/min

per 1M input tokens

Output Token Rate

$0.00

per 1M generated tokens

Context Window

Streaming

maximum token buffer

Avg Throughput

150ms

tokens/sec (typical speed)

Billing Dimension	Standard Rate	Cached / Optimized Rate	Batch (50% Off)	Context Window
Input Prompt	$0.0043/min / M	$0.00 / M	50% discount (24h SLA)	Streaming tokens
Output Generation	$0.00 / M	Standard metered	50% discount (24h SLA)	Up to max model limit
Tool / Function Calling	Standard input tokens	Eligible for prefix cache	Batch supported	Shared context budget

Interactive Monthly Cost Estimator

Estimate your total bill based on prompt size, response volume, and request volume.

Monthly Requests

Avg Input Tokens per Request

Avg Output Tokens per Request

Prompt Cache Hit Ratio (%)

Estimated Monthly Bill

$0.00

Calculated using live 2026 rates for Deepgram Nova 3

Architectural Breakdown & Cost Optimization

Deepgram Nova 3 represents a compelling value proposition in the 2026 developer ecosystem, balancing high accuracy with predictable operating margins. When designing agentic workflows or production RAG systems around Deepgram Nova 3, prompt engineering and caching strategy represent the single largest levers on your monthly bill.

To maximize unit economics:

Leverage Prefix Caching: Structure your system prompts and tool definitions at the beginning of the context payload to guarantee cache hits across multi-turn interactions.
Implement Batch Processing: For non-realtime operations such as data evaluation, document parsing, and backfill tasks, use asynchronous Batch APIs to immediately slash input and output bills in half.
Monitor Token Bloat: Uncontrolled conversational turns degrade both latency and token economics. Prune older history or summarize intermediate thoughts before context exceeds 32k tokens.

Frequently Asked Questions

What is the token pricing for Deepgram Nova 3 in 2026?

Deepgram Nova 3 is billed at $0.0043/min per million input tokens and $0.00 per million output tokens under standard API tiers.

What is the context window capacity of Deepgram Nova 3?

Deepgram Nova 3 supports up to Streaming tokens, allowing long-document ingestion and extensive conversational history.

Does Deepgram Nova 3 support prompt caching or batch discounts?

Yes, prompt caching reduces recurring prefix token costs to roughly $0.00 per million tokens for cached blocks.

Related Frontier Showdowns & Hubs

Deepgram Nova-3 Architecture: End-to-End ASR & Real-Time Streaming TCO

Deepgram's Nova-3 is a cutting-edge speech-to-text (STT) model that sets new industry benchmarks in Word Error Rate (WER) and streaming latency. Engineered specifically for real-time contact centers and conversational voice bots, Nova-3 achieves a 150ms to 200ms preliminary word latency. At $0.0043 per minute for pre-recorded audio and $0.0059 per minute for live streaming (with volume discounts dropping below $0.0035/min), Nova-3 is 4× cheaper and 3× faster than Whisper Large v3.

Speech-to-Text Model	Streaming Rate / Min	Pre-Recorded Rate / Min	Word Error Rate (WER)	Cost per 100,000 Mins
Deepgram Nova-3	$0.0059 / min	$0.0043 / min	6.2% (Industry Benchmark)	$590.00 / month
Whisper Large v3 (OpenAI API)	No native live streaming	$0.0060 / min	7.8%	$600.00 (Batch only)
Google Cloud Speech-to-Text	$0.0160 / min	$0.0160 / min	9.4%	$1,600.00 (2.7× higher)

Production Pipeline: Live Audio WebSocket Streamer in Python

Streaming live microphone or telephony audio to Deepgram Nova-3 with smart formatting and interim transcripts:

from deepgram import DeepgramClient, LiveTranscriptionEvents, LiveOptions

deepgram = DeepgramClient("your-deepgram-api-key")
dg_connection = deepgram.listen.websocket.v("1")

def on_message(self, result, **kwargs):
    transcript = result.channel.alternatives[0].transcript
    if len(transcript) > 0:
        is_final = result.is_final
        print(f"[{'FINAL' if is_final else 'INTERIM'}] {transcript}")

dg_connection.on(LiveTranscriptionEvents.Transcript, on_message)

options = LiveOptions(
    model="nova-3",
    language="en",
    smart_format=True,
    interim_results=True,
    punctuate=True,
    encoding="linear16",
    sample_rate=16000
)

dg_connection.start(options)

Production Best Practices & Audio Optimization

Enable Smart Formatting: Setting `smart_format=True` automatically formats dates, currencies, phone numbers, and email addresses, drastically reducing downstream LLM regex parsing errors.
Multi-Channel Diarization: For call centers recording agent and customer on separate stereo channels, pass `multichannel=true` to transcribe both speakers simultaneously with 100% channel attribution certainty.
Keep-Alive Ping Management: Send keep-alive JSON packets (`{"type": "KeepAlive"}`) over the WebSocket during prolonged user silences to prevent gateway timeout disconnects.

Streaming Speech-to-Text Concurrency & Word Error Rate (WER) Benchmarks

Deepgram Nova-3 represents the state of the art in high-throughput enterprise speech recognition, delivering an industry-leading Word Error Rate (WER) under 6.8% across complex acoustic environments, background babble, and noisy telephony audio streams (8kHz mulaw). From an infrastructure perspective, Nova-3 processes real-time audio at over 40x real-time speed in batch mode and maintains a streaming latency under 200 milliseconds over persistent bidirectional WebSockets.

Enterprise STT Models Accuracy & Cost Comparison (10,000 Hours)

ASR Engine	Telephony WER	Streaming Latency	Cost / Audio Hour	Spend (10,000 Hours)
Deepgram Nova-3	6.8%	180 ms	$0.258 / hr	$2,580.00
OpenAI Whisper Large v3	8.4%	Batch Only (>1,200ms)	$0.360 / hr	$3,600.00
Google Cloud Speech-to-Text V2	9.1%	350 ms	$0.960 / hr	$9,600.00

Live WebSocket Transcription Stream in Node.js

Connect to Deepgram's live transcription WebSocket with automatic endpointing and punctuation:

import { createClient, LiveTranscriptionEvents } from "@deepgram/sdk";

const deepgram = createClient(process.env.DEEPGRAM_API_KEY);
const connection = deepgram.listen.live({
  model: "nova-3",
  punctuate: true,
  interim_results: true,
  smart_format: true
});

connection.on(LiveTranscriptionEvents.Transcript, (data) => {
  const transcript = data.channel.alternatives[0].transcript;
  if (transcript) console.log(`[Nova-3] ${transcript}`);
});

Deepgram Nova-3 High Concurrency Call Center SLAs

For enterprise contact centers routing thousands of concurrent telephone interactions, Nova-3 guarantees an 99.99% service availability SLA with a median streaming latency under 180 milliseconds. Audio streams that experience unexpected socket disconnects benefit from automated connection reconnection tokens that resume transcription without losing prior speaker diarization state.