Can Flexprice Handle Cohere's Pricing?

Cohere prices Command R production API usage at $0.15 per million input tokens** and **$0.60 per million output tokens. Its response distinguishes billable tokens from total processed tokens. This guide maps the billed quantities to two Flexprice meters and a usage-only plan for command-r-08-2024.

Cohere Generative models pricing comparison showing Command R input at 0.15 per million tokens and output at 0.60 per million tokens beside the complete Command A+, Command R7B and Transcribe cards — Cohere's complete Generative models comparison, captured on November 11, 2026, with all four cards, token rates, inclusions and the API-key entry point visible. Only Command R production API token charges are modeled; Command A+, Command R7B and Transcribe are shown for context.

Source: Cohere pricing, with Generative models selected.

This example covers USD token charges on Cohere’s direct API using a production key. It excludes trial-key usage, other models and endpoints, private or dedicated deployments, partner-hosted pricing, discounts and tax. The illustrative Flexprice plan bills monthly in arrears. It does not reproduce Cohere’s additional invoice trigger at $250 of outstanding balance.

Pricing profile

These labels describe the Command R production API offering, not every Cohere product or sales channel.

Dimension	Offering profile
Pricing model	Usage-based
Industry	AI language models
Buyer	Developers
Sales motion	Self-serve signup; Production activation required
Billing mechanics	Separate input/output rates; Billed-token metering; Monthly and balance-threshold invoicing

The pricing page offers API-key signup. Its FAQ says an organization owner requests a production key through the billing workflow. Trial access is not permission to use the free key for production or commercial workloads.

Command R token prices

The Generative models rate cards publish the following prices. The Command R documentation identifies the August 2024 API model as command-r-08-2024. Do not apply these rates to the older March 2024 model listed separately in the pricing FAQ.

Billable quantity	USD per million tokens	Equivalent USD per token
Input tokens	$0.15	$0.00000015
Output tokens	$0.60	$0.0000006

One million tokens is the rate’s denominator, not a minimum package. Track raw integer counts of billed tokens and retain the decimal per-token prices. Cohere’s billing guide distinguishes billed_units from tokens. Cohere can add tokens internally that are not charged to the customer. For the v2 Chat response, use usage.billed_units.input_tokens and usage.billed_units.output_tokens, not the corresponding totals under usage.tokens.

Design takeaways

Meter the billing measure, not the closest telemetry measure. Cohere’s billed-versus-total distinction means a generic token counter can overstate charges. Normalize the provider’s billed fields before sending usage to Flexprice.
Keep input and output independently priced. The output rate is four times the input rate. Two SUM meters preserve the mix instead of hiding it inside a blended rate or one total-token count.
Separate rate calculation from invoice timing. Cohere’s month-end or $250 balance trigger changes when collection occurs, not the token unit prices. A monthly plan can model the usage charges without claiming to recreate the extra invoice trigger.

How Cohere’s pricing maps to Flexprice

Cohere’s pricing uses	Flexprice setup
A named model with its own rate card	A `model` property and a `command-r-08-2024` filter on both metered features
Production-key usage, not free trial calls	A normalized `key_type = production` filter on both meters
Billed input and output tokens	Separate SUM meters over `billed_input_tokens` and `billed_output_tokens`
Two constant per-token rates	Two flat-fee usage charges, each linked to its own meter
Monthly invoice plus a balance trigger	Monthly arrears for this example; balance-triggered invoicing remains outside scope

The event and plan names below are integration-defined examples, not Cohere API resources. Do not add a fixed subscription fee, token allowance or credit grant to this usage-only example. Do not invent a separate cache rate from another provider’s pricing model.

Set up the metered features

Create an input-token feature. Both filters must match for an event to contribute usage:

{
  "name": "Cohere Command R billed input tokens",
  "lookup_key": "cohere_command_r_billed_input_tokens",
  "type": "metered",
  "unit_singular": "token",
  "unit_plural": "tokens",
  "meter": {
    "name": "Cohere Command R billed input tokens",
    "event_name": "cohere_chat_usage",
    "reset_usage": "BILLING_PERIOD",
    "aggregation": { "type": "SUM", "field": "billed_input_tokens" },
    "filters": [
      { "key": "model", "values": ["command-r-08-2024"] },
      { "key": "key_type", "values": ["production"] }
    ]
  }
}

Create the output-token feature using the same event name and filters, but a different aggregation field:

{
  "name": "Cohere Command R billed output tokens",
  "lookup_key": "cohere_command_r_billed_output_tokens",
  "type": "metered",
  "unit_singular": "token",
  "unit_plural": "tokens",
  "meter": {
    "name": "Cohere Command R billed output tokens",
    "event_name": "cohere_chat_usage",
    "reset_usage": "BILLING_PERIOD",
    "aggregation": { "type": "SUM", "field": "billed_output_tokens" },
    "filters": [
      { "key": "model", "values": ["command-r-08-2024"] },
      { "key": "key_type", "values": ["production"] }
    ]
  }
}

Capture each returned meter ID separately. Prices reference meter IDs, not feature IDs. Leave reporting-unit conversion unset so the meters retain raw token counts.

Configure the usage-only plan

Create a plan without a recurring fixed charge:

{
  "name": "Cohere Command R usage example",
  "lookup_key": "cohere_command_r_usage_example",
  "description": "Billed input and output token charges only; monthly arrears"
}

Add the input price, replacing the example plan and meter IDs with the returned values:

{
  "entity_type": "PLAN",
  "entity_id": "plan_cohere_command_r_example",
  "meter_id": "meter_cohere_command_r_input_example",
  "type": "USAGE",
  "price_unit_type": "FIAT",
  "billing_model": "FLAT_FEE",
  "amount": "0.00000015",
  "currency": "usd",
  "billing_period": "MONTHLY",
  "billing_period_count": 1,
  "invoice_cadence": "ARREAR",
  "display_name": "Command R billed input tokens"
}

Add the output price to the same plan with its own meter:

{
  "entity_type": "PLAN",
  "entity_id": "plan_cohere_command_r_example",
  "meter_id": "meter_cohere_command_r_output_example",
  "type": "USAGE",
  "price_unit_type": "FIAT",
  "billing_model": "FLAT_FEE",
  "amount": "0.0000006",
  "currency": "usd",
  "billing_period": "MONTHLY",
  "billing_period_count": 1,
  "invoice_cadence": "ARREAR",
  "display_name": "Command R billed output tokens"
}

FLAT_FEE here means a constant usage rate, not a fixed monthly subscription charge. Do not use a package that rounds usage up to a million tokens. Add an unlimited metered entitlement for each feature, then create a customer and a subscription to this plan with quantity 1. Unlimited access does not make usage free.

Normalize and send usage

Read the completed provider response on your server. Map the fields as follows:

Source	Normalized Flexprice property
Model ID used in the request	`model`
Whether the credential is a production key	`key_type`
`usage.billed_units.input_tokens`	`billed_input_tokens`
`usage.billed_units.output_tokens`	`billed_output_tokens`

Do not send the API key itself. Your integration supplies the non-sensitive key_type label; it is not a Cohere response field. Do not substitute usage.tokens if billed quantities are missing. Hold those records for reconciliation instead of silently guessing the charge. For a completed production call with 6,772 billed input tokens and 248 billed output tokens, the normalized usage event is:

{
  "event_name": "cohere_chat_usage",
  "event_id": "evt_cohere_command_r_0001",
  "external_customer_id": "customer_ai_app_001",
  "timestamp": "2026-11-11T04:00:00Z",
  "properties": {
    "model": "command-r-08-2024",
    "key_type": "production",
    "billed_input_tokens": "6772",
    "billed_output_tokens": "248"
  }
}

The event uses string-valued dynamic properties. The two quantities contribute $0.0011646 before currency rounding: 6772 × 0.00000015 + 248 × 0.0000006. They illustrate normalization, not a live request executed by this guide. Send the final billed counts once, not both streaming deltas and the final total. Use a stable event ID and preserve the original timestamp when retrying delivery of the same usage record. Flexprice deduplication requires both to match. A new separately charged provider call is a different record. Do not send a batch together with the individual records it contains.

Worked usage calculation

Suppose one customer’s billing period contains 40 million billed input tokens and 10 million billed output tokens, all for command-r-08-2024 using production keys:

Component	Calculation	Charge
Billed input	40,000,000 × $0.00000015	$6.00
Billed output	10,000,000 × $0.0000006	$6.00
Usage total	$6.00 + $6.00	$12.00

The same 50 million total billed tokens would cost a different amount with a different input/output mix. If all 50 million were input, the charge would be $7.50; if all were output, it would be $30.00. Keep both quantities separate through aggregation. Retain decimal precision until invoice currency rounding. Do not round every request to cents or round token quantities up to a million. The $12.00 excludes discounts and tax. It demonstrates the usage-rate calculation, not Cohere’s full payment workflow or a guarantee about production-key approval. This guide does not execute setup or API calls.

Pricing sources

Cohere pricing: current Generative models rates, trial/production distinction, production activation and invoice timing.
How Cohere pricing works: separate input/output prices and billed-versus-total tokens.
Cohere v2 Chat API: model selection and response usage structure.
Command R: the command-r-08-2024 model identifier and offering scope.

Pricing checked on 2026-11-11. For the underlying setup workflow, see the usage-based pricing recipe. For a separate prepaid settlement model, see the credit-based pricing recipe.

Billing Recipes

Pricing Teardowns

Can Flexprice Handle Cohere's Pricing?

Pricing profile

Command R token prices

Design takeaways

How Cohere’s pricing maps to Flexprice

Set up the metered features

Configure the usage-only plan

Normalize and send usage

Worked usage calculation

Pricing sources

​Pricing profile

​Command R token prices

​Design takeaways

​How Cohere’s pricing maps to Flexprice

​Set up the metered features

​Configure the usage-only plan

​Normalize and send usage

​Worked usage calculation

​Pricing sources

Pricing profile

Command R token prices

Design takeaways

How Cohere’s pricing maps to Flexprice

Set up the metered features

Configure the usage-only plan

Normalize and send usage

Worked usage calculation

Pricing sources