Which AI vendors actually
let you track what you spend?

Per-model, per-token, per-invocation; plus per-IAM-principal, per-application-inference-profile, per-project/workspace, and per-request-metadata attribution (native dollar grain is per usage type per day)

Per-model, per-token, per-deployment, subscription-level

Per-model, per-prediction, SKU-level

Cost/usage by billing-item, instance, model, API key, workspace, token type; daily or monthly cycle

per-App, per-day or per-hour cost; optional per-App-tag breakdown

Per-model, per-request, tokens aggregated to API key*

Per-model, per-workspace, per-API-key, per-service-tier, per-context-window; aggregated in 1m/1h/1d buckets, per-request tokens in the response usage object

Per-request and per-billing-event cost in USD; aggregated usage by timeframe (minute/hour/day/week/month); on-demand FOCUS CSV report

per-request (POST /v1/request-costs, costNanoUsd) and per-model-per-month aggregate (GET /payment/usage)

per-bill line with cycleType (Hour/Day/Week/Month) and productCategory (llm/gpu/storage); monthly bills; USD as 1/10000 USD

per serverless-endpoint / pod / GPU-type, time-bucketed (hour to year) over an arbitrary startTime/endTime range

SKU and character-transaction level detail

Per-model, per-request tokens

Per-model, per-request (input tokens, output tokens, search units)

per-request token counts in response; per-currency balance via GET /user/balance; manual per-month export

per-request response usage; daily buckets (billingUsage) grouped by model/api_key/deployment/tags; per-event export-metrics CSV (tokens-only)

per-request token counts in response; account balance in USD; aggregated usage/cost dashboard-only

Per-character/credit, per-product, per-voice, per-API-key, per-user, per-model, per-region, per-workspace; USD spend via fiat_units_spent metric

hourly FOCUS 1.2 line items with per-resource attribution over a billing-period range; Token Factory per-request token counts

per-request token usage in response; dashboard aggregates by model

per-request response usage (cost_in_usd_ticks, 1 USD = 1e10 ticks) plus aggregated history via Management API /usage with time buckets, group_by, and aggregations

Per-model, per-request tokens

Compute seconds per run; multiply by hardware rate for USD

Per-deployment / per-model / per-training-job; daily; tokens and billable minutes

Per-generation cost object (amount plus unit, dollars or credits) in response; USD account balance under PAYG; credit balance via GET /me

Credit balance and 90-day usage history by model and day

Per-request USD cost in response (opt-in); account balance and aggregated usage (credits/requests) via accountManagement getDetails for today/7d/30d/lifetime

Credit balance via API, per-operation dollar costs

Credit balance via billing endpoint; credits require rate conversion

Credit balance via API (returned in USD cents); per-generation credit costs; queryable generations list endpoint

Per-request cost in response, credit balance via API

Per-request, per-model, per-user token and cost data

Per-request token counts in API response; aggregated token usage queryable by day/hour via GetUsage API, by project/endpoint

Per-request token counts in response only

per-request token usage in response; Prometheus per-minute metrics; Console cost by model/token type per month

Per-request token counts; current-window quota-remaining endpoint

Index-level metrics; ecommerce event analytics (search/click/conversion)

None

None