Can Flexprice Handle Deepgram's Pricing?

Deepgram prices speech-to-text by model, processing mode and audio duration. Its rate table displays prices per minute, but normal audio requests are billed per second without rounding up to a minute. This guide models Flux STT Multilingual streaming on Pay as you go with one metered feature and one usage charge.

Deepgram Pay as you go, Growth and Enterprise plan cards showing no minimum for Pay as you go — Deepgram plan cards, captured from its official pricing page on November 4, 2026. This teardown uses Pay as you go base usage pricing only; the visible promotional credit, Growth commitments and Enterprise agreements are excluded.

Deepgram streaming STT rate table with Flux STT Multilingual at $0.0078 per minute on Pay as you go, and a note that normal audio requests are billed per second without block rounding — Deepgram streaming speech-to-text rates, captured on November 4, 2026. Only Flux STT Multilingual on Pay as you go at $0.0078 per minute is modeled, billed per second; other models, plans and multichannel audio are excluded.

Source: Deepgram’s official pricing page.

This example covers hosted, mono audio using flux-general-multi, with English and Spanish language hints, in USD. It excludes pre-recorded or batch transcription, the English-only Flux model, Nova models, multichannel audio, paid add-ons, Audio Intelligence, text-to-speech, Voice Agent bundles, self-hosting, promotional credits, Growth commitments and Enterprise agreements. It also excludes request options that change the published base rate. The configuration names below are examples, not Deepgram product identifiers.

Pricing profile

These labels describe only the mono Flux Multilingual Pay as you go streaming offering modeled here, not Deepgram as a whole.

Dimension	Offering profile
Pricing model	Usage-based
Industry	AI
Buyer	Developers
Sales motion	Self-serve
Billing mechanics	Per-second metering; No duration minimum; No block rounding

Flux Multilingual streaming prices

On the official pricing page, select Speech-to-text, then Streaming. Use the Flux STT Multilingual row and the Pay as you go column, not a pre-recorded or English-only rate.

Pricing dimension	Selected scope
Published rate	$0.0078 per minute of processed audio
Billing unit	Seconds of processed audio
Equivalent price per second	$0.0078 ÷ 60 = $0.00013
Duration minimum	None for normal audio requests
Block rounding	None; do not round each request up to a minute
Plan commitment	None on Pay as you go

The price is for the multilingual model, not the detected language of each turn. Supplying only an English hint does not select the separately priced English-only model. The Flux language guide documents flux-general-multi, its ten supported languages and optional language hints. For this example, your application connects to the hosted streaming endpoint documented in the Flux quickstart:

wss://api.deepgram.com/v2/listen?model=flux-general-multi&language_hint=en&language_hint=es&encoding=linear16&sample_rate=16000

This URL assumes mono, raw 16-bit PCM audio at 16 kHz. It is not the /v1/listen pre-recorded endpoint. Language hints bias recognition; they do not restrict all speech to those languages.

How Deepgram’s pricing maps to Flexprice

Deepgram’s pricing uses	Flexprice setup
A rate specific to the model and streaming mode	A metered feature filtered to `model = flux-general-multi`, `mode = streaming` and the selected mono base-rate scope
Processed audio seconds	A SUM meter over the normalized `billable_seconds` property, not COUNT of requests or transcript turns
A displayed per-minute rate without minute blocks	A flat-fee usage charge of `0.00013` USD per raw second, without a package transform
No duration minimum	No minimum quantity or minimum-spend commitment on this charge
No included monthly minutes in this example	An unlimited metered entitlement for access, with every recorded second priced by the usage charge

The example uses monthly usage billing in arrears in Flexprice. That is an implementation choice, not Deepgram’s payment cadence: Deepgram says Pay as you go usage draws down credits. Its advertised $200 starting credit is outside this example and must not become a recurring monthly free-minute allowance. Model any prepaid funding separately with wallet credits.

Set up the seconds meter

In Product Catalog → Features → Add Feature, create a metered feature named Flux Multilingual streaming seconds. Use these settings, or the equivalent feature configuration through the API:

{
  "name": "Flux Multilingual streaming seconds",
  "lookup_key": "flux_multilingual_streaming_seconds",
  "type": "metered",
  "unit_singular": "second",
  "unit_plural": "seconds",
  "meter": {
    "name": "Flux Multilingual streaming seconds",
    "event_name": "deepgram_audio_processed",
    "reset_usage": "BILLING_PERIOD",
    "aggregation": {
      "type": "SUM",
      "field": "billable_seconds"
    },
    "filters": [
      { "key": "model", "values": ["flux-general-multi"] },
      { "key": "mode", "values": ["streaming"] },
      { "key": "audio_scope", "values": ["mono_base"] }
    ]
  },
  "reporting_unit": {
    "unit_singular": "minute",
    "unit_plural": "minutes",
    "conversion_rate": 60
  }
}

The event properties and mono_base label are your integration’s conventions. Only label a request mono_base after verifying that it matches the mono base-rate scope, without paid extras. Route other models, modes and options to their own meters and prices. The optional reporting unit displays raw seconds ÷ 60 as minutes. For example, 6,000 seconds displays as 100 minutes. It does not change the SUM meter or the per-second price. Do not divide usage by 60 again before applying that price.

Configure the usage-only plan

Create a plan named Flux Multilingual PAYG Example. Add the feature as a usage-based flat-fee charge with the following configuration:

Setting	Value
Currency / price unit type	`usd` / `FIAT`
Charge type / billing model	`USAGE` / `FLAT_FEE`
Amount per second	Decimal string `"0.00013"`
Billing period / count	`MONTHLY` / `1`
Invoice cadence	`ARREAR`
Meter	The meter attached to the feature created above
Minimum quantity, tiers and package transform	Omit; no minimum or block rounding
Recurring fee or spend commitment	None

When using the prices API, set entity_type to PLAN, entity_id to the created plan’s ID and meter_id to the returned meter ID. Use billing_period, not a differently named interval field. Preserve the decimal amount rather than rounding the per-second rate to cents. In the plan’s Entitlements tab, add the feature with unlimited metered access (Set to Infinite). This permits access without a usage cap; it does not make usage free. Do not add a free-usage tier, recurring credit grant or minimum-spend commitment to this scoped example.

Send usage and verify the arithmetic

Subscribe a customer to the plan and send usage events from your server. Normalize each completed streaming request into one event, such as this illustrative 14-second request:

{
  "event_name": "deepgram_audio_processed",
  "external_customer_id": "customer_example",
  "event_id": "deepgram_request_example_001",
  "timestamp": "2026-11-04T00:00:00Z",
  "properties": {
    "model": "flux-general-multi",
    "mode": "streaming",
    "audio_scope": "mono_base",
    "billable_seconds": "14"
  }
}

billable_seconds is a normalized application property, not a claimed Flux response field. Supply the processed-audio duration from your integration and reconcile it with Deepgram’s request usage information before billing. Do not infer billable duration from transcript word timestamps or use WebSocket connection time, which the pricing page specifies for the separate Voice Agent product. This example assumes the durations below are already verified billable seconds; it does not define an additional duration-rounding rule. Keep a stable event_id derived from the provider request identifier when retrying event delivery. Emit the final usage once per request, not once per interim transcript, final turn and session total. A new upstream request is distinct from retrying delivery of the same usage event. Use decimal arithmetic to check both a short request and a monthly aggregate of 1,000 requests at 14 seconds plus 500 requests at 92 seconds:

from decimal import Decimal

per_minute = Decimal("0.0078")
per_second = per_minute / Decimal("60")
assert per_second == Decimal("0.00013")

short_request = Decimal("14") * per_second
assert short_request == Decimal("0.00182")

total_seconds = Decimal("1000") * Decimal("14") + Decimal("500") * Decimal("92")
assert total_seconds == Decimal("60000")
assert total_seconds / Decimal("60") == Decimal("1000")
assert total_seconds * per_second == Decimal("7.80")

Usage	Processed seconds	Calculation	USD charge before invoice rounding
One short request	14	14 × 0.00013	$0.00182
1,000 short requests	14,000	14,000 × 0.00013	$1.82
500 longer requests	46,000	46,000 × 0.00013	$5.98
Monthly total, excluding the separate single-request check	60,000	60,000 × 0.00013	$7.80

Do not round the $0.00182 short-request charge to cents before aggregation. A 14-second request is not a full-minute package. Compare the meter’s raw seconds, displayed minutes and usage charge against these values before enabling live billing. This guide describes a configuration; no live resources or invoices have been created.

Design takeaways

These are modeling interpretations of the sourced mechanics, not claims about Deepgram’s intent.

Keep display units out of the billing calculation. The pricing page presents minute rates but bills normal audio requests per second without block rounding. Here that means summing raw seconds and applying an exact per-second charge; reporting minutes should not introduce a package boundary.
Route usage by requested model, not detected language. Language hints operate within the multilingual model, while the rate table prices Flux English and Multilingual separately. Filter the Flexprice meter by requested model and streaming scope so English speech processed by flux-general-multi cannot accidentally receive the English-only rate.

Pricing sources

Pricing checked on 2026-11-04.

Deepgram pricing: Speech-to-text → Streaming → Flux STT Multilingual → Pay as you go rate; per-second billing, no duration minimum or block rounding, credit-based payment and starting credit.
Flux quickstart: hosted /v2/listen endpoint, model selection and raw audio format requirements.
Flux Multilingual and Language Prompting: supported languages and English/Spanish hint configuration.
Get a Project Request: request-level usage information for reconciliation.
Deepgram speech-to-text product guide: conversational AI use cases, Flux model positioning and the developer audience for the API.

Billing Recipes

Pricing Teardowns

Can Flexprice Handle Deepgram's Pricing?

Pricing profile

Flux Multilingual streaming prices

How Deepgram’s pricing maps to Flexprice

Set up the seconds meter

Configure the usage-only plan

Send usage and verify the arithmetic

Design takeaways

Pricing sources

Build Usage-Based Pricing

​Pricing profile

​Flux Multilingual streaming prices

​How Deepgram’s pricing maps to Flexprice

​Set up the seconds meter

​Configure the usage-only plan

​Send usage and verify the arithmetic

​Design takeaways

​Pricing sources

​Related recipe

Build Usage-Based Pricing

Pricing profile

Flux Multilingual streaming prices

How Deepgram’s pricing maps to Flexprice

Set up the seconds meter

Configure the usage-only plan

Send usage and verify the arithmetic

Design takeaways

Pricing sources

Related recipe