In an agentic session, thinking blocks from previous turns are retained by the API and re-sent as input tokens on every subsequent turn. The cost compounds, not accumulates.
By Pawan Nayar · from the LLOS AI Learning record
In an agentic session, thinking blocks from previous turns are retained by the API and re-sent as input tokens on every subsequent turn.
The cost compounds, not accumulates.
The API does not discard thinking blocks after a turn ends — it keeps them as part of the conversation history. Every new turn re-bills all prior thinking as input tokens. A debugging conversation on turn 20 pays the full thinking cost of turns 1 through 19, again. The per-token rate does not change, but the volume multiplies with every turn. A thinking-heavy session on turn 30 can cost ten times what the isolated turn count implies. This is the default contract — not a billing error.
A developer debugs an output-truncation issue across 25 turns with thinking enabled. Turn 25 re-sends all thinking from turns 1–24 as input. The session bill is roughly 12× what 25 isolated turns would cost at face value.
Resolved from this lesson's own record — real definitions, not invented for this page.
More from the same record, pulled by shared topic — not written new for this page.
An agentic framework like Claude Code holds conversation state, but the API has no concept of already-paid-for. Each turn packages the full history and sends it as input.
In an agentic session, thinking blocks from previous turns are retained by the API and re-sent as input tokens on every subsequent turn. The cost compounds, not accumulates.