# Arize Phoenix now updates built-in model prices automatically

published: 2026-11-02T15:05:24.999Z
topic: infrastructure
source: https://arize.com/blog/phoenix-automatic-model-pricing-updates/
url: https://aiagentnews.io/stories/arize-phoenix-now-updates-built-in-model-prices-automatically

Arize Phoenix now refreshes its built-in model pricing between releases, so new or revised rates reach a server without an upgrade. Servers check for a newer catalog shortly after startup and about once a day, applying it in the background.

Phoenix uses a pricing catalog to calculate the cost of each LLM call in a trace by applying the model's token rates to the recorded token usage. Previously, built-in price changes arrived only with new Phoenix releases, and a model missing from the catalog required an upgrade or a custom configuration.

The server downloads the catalog from the Arize Phoenix repository on GitHub, which Arize generates from LiteLLM's model price list. If the downloaded catalog is newer than the one in use, Phoenix saves the model entries and token prices together in one database transaction. Failed downloads leave existing prices in place, and the server retries at the next scheduled check.

Updated rates apply to new spans after the pricing cache refreshes, and costs already recorded on existing traces are not recalculated. Custom prices continue to override built-in prices. Automatic updates are on by default and can be turned off with PHOENIX_DISABLE_MODEL_COST_AUTO_UPDATE=true. The Models page under Settings shows the date of the catalog in use.

## Why it matters if you build agents

Teams running self-hosted Phoenix no longer need a release upgrade to pick up new model rates, so cost figures in traces can track provider pricing more closely.

Costs on existing traces will not change when prices update, so comparisons across time should account for which rates applied to each span. Teams that review pricing as part of their release process can set the disable variable to keep control of updates.