Model Usage and Cost Management

This feature is EXPERIMENTAL and under active development. It may change significantly, be discontinued, or have breaking changes without notice. Documentation may be incomplete or outdated and is NOT recommended for production use. Use at your own risk. Please refer to our Upgrade and Release Process for more information.

Overview

The Model Usage and Cost Management feature gives organisations full visibility into how their users interact with AI models — tracking token consumption and estimated spend per user, assistant, and application in real time.

Once enabled, the feature:

Records every LLM call (input tokens + completion tokens + estimated cost in USD)
Shows an Admin dashboard with usage reporting and model pricing
Shows a per-user spend badge in the Chat interface
Supports CSV export for offline analysis

Enabling the Feature

Model Usage tracking is controlled by two feature flags. Both are disabled by default and must be activated by Unique.

Feature Flag	What it enables	Feature Status

Feature Flag	What it enables	Feature Status
`FEATURE_FLAG_SAVE_MODEL_USAGE_UN_12832`	Starts recording usage data to the database. Must be on for anything to work.	Experimental
`FEATURE_FLAG_SAVE_MODEL_USAGE_DASHBOARD_UN_18889`	Shows the Cost Management section in the Admin UI and the spend badge in the Chat user menu.	Experimental

To enable these flags for a client, please contact your Unique Representative or raise a request via the Enterprise mailbox.

Admin UI — Cost Management

Visible in the Settings Page → Cost Management (left sidebar), once the dashboard flag is enabled.

Requires the user to have the role CHAT_FEEDBACK_READ or CHAT_DATA_ADMIN.

Model Pricing

A read-only table showing the active pricing for every supported model:

Column	Description

Column	Description
Model	Internal model identifier (e.g. `AZURE_GPT_4o_2024_1120`, `litellm:anthropic-claude-sonnet-4-5`)
Input cost / 1M tokens	Cost in USD per million prompt tokens
Completion cost / 1M tokens	Cost in USD per million completion (output) tokens
Currency	USD by default

Usage Reporting

An aggregated view of token consumption and estimated spend across the organisation.

Filters available:

Date range — This month, last month, this week, or custom range (with previous/next period navigation)
View by — Model, User, or Assistant
Drill-down — Click any row to see the breakdown within that dimension
Text filter — Search by model name, user, or assistant
Pagination — Configurable page size

How Costs Are Calculated

Cost is computed per LLM call using the formula:

spent = (inputTokens x inputCostPer1M + completionTokens x completionCostPer1M) / 1,000,000

All costs are expressed in USD.

Default Pricing Sources

Default prices are maintained by the Unique engineering team and sourced from:

Azure OpenAI models — Azure OpenAI Service Pricing (Sweden Region): https://azure.microsoft.com/en-us/pricing/details/azure-openai/
LiteLLM-proxied models (Anthropic Claude, Gemini, Mistral, OpenAI via LiteLLM) — LiteLLM Providers and Models: https://models.litellm.ai/

Sample Default Prices (April 2026)

Model	Input (USD / 1M tokens)	Completion (USD / 1M tokens)

Model	Input (USD / 1M tokens)	Completion (USD / 1M tokens)
AZURE_GPT_4o_2024_1120	$2.50	$10.00
AZURE_GPT_4o_MINI_2024_0718	$0.15	$0.60
AZURE_GPT_41_2025_0414	$2.00	$8.00
AZURE_o3_2025_0416	$2.00	$8.00
AZURE_o4_MINI_2025_0416	$1.10	$4.40
litellm:anthropic-claude-sonnet-4-5	$3.00	$15.00
litellm:anthropic-claude-opus-4-5	$5.00	$25.00
litellm:gemini-2-5-pro	$1.25	$10.00

Per-Client Price Override

Prices can be customised per client. If a client's Azure contract or LiteLLM agreement carries different rates, the Unique engineering team can supply a custom pricing configuration for that deployment. Contact your CS to arrange this.

CSV Export

Usage data can be exported as a CSV file via the analytics export pipeline. User privacy settings (pseudonymisation or anonymisation) are respected per organisation configuration.

CSV columns:

Column	Description

Column	Description
S/N	Row sequence number
User ID	User identifier (may be pseudonymised/anonymised per org settings)
Assistant ID	The assistant used (N/A if not applicable)
Chat ID	The conversation session
App ID	The Unique application
Language Model	Model identifier
Spent	Estimated cost in USD
Input Tokens	Number of prompt tokens
Completion Tokens	Number of output tokens
Timestamp	UTC timestamp of the LLM call

Required User Roles

Role	What it grants

Role	What it grants
`CHAT_FEEDBACK_READ`	View usage reporting and model pricing in Admin
`CHAT_DATA_ADMIN`	Full access to model usage data in Admin