Conversation token cost

The conversation token cost panel gives you per-conversation visibility into LLM token usage and estimated costs. It appears in the conversation details view inside Studio > Conversation logs, letting you understand exactly how many tokens each conversation consumed and how much it cost.

What you can see

When you open a conversation and switch to the token cost tab, the panel shows aggregate statistics for the entire conversation: total prompt tokens, completion tokens, cached tokens, total LLM calls, and the estimated cost in USD.

Below the aggregates, a per-model breakdown lists every model and provider that was invoked during the conversation, along with its individual token counts and cost. This helps you identify which models are driving spend.

A chronological timeline of individual LLM calls is also available. Each entry shows the timestamp, model, provider, feature that triggered the call, execution time, token counts, and estimated cost. This is useful for debugging slow or expensive conversations.

Feature gating

The token cost panel is gated behind a feature flag. When the flag is not enabled for your account, the tab does not appear. If you are interested in enabling this feature, contact your account administrator.

Cost estimation

Costs are estimated based on published per-token pricing for each model and provider. The values shown are approximations and may not exactly match your invoice, since pricing tiers, discounts, and cached-token rates can vary by account.