What's New

This release brings cost visibility to Aiceberg. You can set per-model pricing and see input, output, and total cost across monitoring and analytics, alongside a new profile version indicator and a timestamp correction.

Per-Model Pricing

You can now set pricing for a model when you create it, in the Pricing section on the model page. Pricing takes an Input Cost and an Output Cost, each expressed in USD per one million tokens. Input Cost covers the prompt tokens sent to the model. Output Cost covers the completion tokens it returns. Both values are optional. Once set, pricing appears on the model detail page.

Pricing is what produces the cost figures now surfaced across monitoring and analytics. A model needs pricing configured before its usage shows a dollar cost anywhere in the platform.

Cost Analytics

Cost visibility is now built into monitoring and analytics. On the Monitoring event table, three new columns report cost per event: Input Cost, Output Cost, and Total Cost. The columns are hidden by default; add them from the column picker.

On Use Case Analysis, a new Cost by LLM Model module breaks down spend per model, with input, output, and total figures for each. It shows which models are driving cost across the use case.

Profile Version in the Detail View

The profile detail page now shows the profile version in the metadata strip, between Created and Lasted edited, so you can see at a glance which version you are looking at.

Bug Fix

Timestamps now display in your correct local time. Previously, some timestamps could appear shifted by your timezone offset. Timestamps are now interpreted consistently as UTC and rendered correctly wherever they appear, including the log drawer.

Did this answer your question?