PulseAugur
EN
LIVE 03:16:48

Claude Code agent consumes 3.57B tokens in one turn, driven by subagents

A single turn of the Claude Code agent consumed an astonishing 3.57 billion tokens, with the majority attributed to subagents and cache reads. The agent, running on Claude Code 2.1.317 and Opus 5.5, involved 22 long-lived 'teammate' subagents that accounted for 73% of the token usage. This high token consumption was primarily due to the agents re-reading extensive conversation history, with the cost of starting new subagents being negligible compared to maintaining active ones. AI

IMPACT Highlights the significant token costs associated with complex, multi-agent AI systems and the importance of efficient context management.

RANK_REASON The item discusses the token consumption of a specific AI agent, which is a detail about the operational use of an AI product rather than a new release or significant industry event.

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Claude Code agent consumes 3.57B tokens in one turn, driven by subagents

How we ranked this

Signal score
49 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The item discusses the token consumption of a specific AI agent, which is a detail about the operational use of an AI product rather than a new release or significant industry event.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
product, infra
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Rulestack ·

    One turn of our Claude Code agent used 3.57 billion tokens. 22 long-lived teammates spent 2.34 billion of it

    <blockquote> <p>One prompt to our Claude Code agent ran for 23 hours and used 3,567,236,645 tokens. Subagents used 90% of them, and 97% of all of it was cache reads. What surprised us was where the subagent share went: 22 long-lived teammates used 73% of it, and 61 short review s…