Skip to main content

FMAPI Tokens

Use Pricing > FMAPI Tokens to compare proprietary Foundation Model API rate combinations loaded in Lakemeter.

The view helps identify the correct model, geography, context-length, and token-rate assumptions before creating FMAPI workload lines. It does not determine how many tokens an application will use.

Model availability and rates change. Use the values shown in Lakemeter for planning and validate current product guidance in the official Databricks documentation and current public pricing on the Databricks pricing page.

FMAPI Tokens pricing view Filter proprietary model rates and display them as DBUs or estimated currency values.

Choose a display mode

DBU

The DBU view shows the source billing rates without applying a regional price per DBU:

  • Token rates are displayed in DBUs per one million tokens.
  • Batch inference rates are displayed in DBUs per hour.

Use this mode to compare relative consumption across model and rate combinations.

Currency

The $ view converts DBU rates to estimated currency values using the selected cloud, region, tier, discount, and FX assumptions:

Displayed token rate
= DBUs per billing unit
× Regional price per DBU
× (1 − Discount percentage)
× FX

The dollar view shows Global and In-geo values side by side when both are available.

Filter the rate matrix

Use the filters to narrow the table:

FilterSizing use
CloudSelect the cloud context for the rate
RegionSelect the regional price per DBU used in currency mode
TierSelect the pricing tier used in currency mode
FamilyLimit the table to a model provider or family
Model containsSearch within available model identifiers
GeoFilter the geography in DBU mode
Context lengthSelect the context-rate category
Discount %Apply a planning discount in currency mode
USD to local FXShared Pricing-page control that converts displayed dollar values with a manually supplied multiplier

The controls change with the selected display mode because the DBU view compares source token rates, while currency mode also needs a regional DBU price. The FX input remains in the Pricing page header and applies across both Pricing tabs.

Read the table

The table separates model context from billable rate types.

Model context

  • Cloud
  • Family
  • Model
  • Geography, where shown
  • Context-length category

Token and batch rates

  • Input tokens
  • Output tokens
  • Cache writes
  • Cache reads
  • Batch inference

A dash means Lakemeter does not have a published rate for that combination. Do not treat a dash as a zero-cost rate.

Turn a rate into a sizing assumption

For token-based usage:

Monthly DBUs
= Monthly tokens in millions
× DBUs per one million tokens

For batch inference:

Monthly DBUs
= Batch hours
× DBUs per hour

Create separate FMAPI workload lines for separately billed rate types such as input tokens, output tokens, cache reads, or cache writes. The Foundation Models — Proprietary sizing guide explains how those quantities are entered in an estimate.

Review checklist

  • Is the model identifier correct?
  • Does the geography match the planned serving option?
  • Is the context-length category appropriate?
  • Are input, output, cache, and batch quantities modeled separately?
  • In currency mode, are region, tier, discount, and FX correct?
  • Have missing rates been treated as unavailable rather than free?