Token Reduction Is Not Cost Reduction
Quick Answer
This paper shows that Reducing token counts in coding agents does not equate to cost savings, as shown in a study of 2,848 Claude Code runs.
Quick Take
The findings reveal that prompt-cache traffic accounts for 87% of costs, while tool-output reduction can actually increase expenses and hinder task completion.
Key Points
- Prompt-cache traffic constitutes 87% of the total costs in coding agent operations.
- Tool-output reduction led to a 6.8% increase in billed costs in some cases.
- Compression can negatively impact task success by removing critical evidence.
- A proposed standard emphasizes success-adjusted billed costs over mere token reduction.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Context-reduction layers for API-based coding agents, including command-output compressors, retrieval rankers, and payload-optimizing proxies, are usually evaluated by how much text they remove. We ask instead: when does reducing retrieved context or tool output lower the actual billed cost of a coding agent without reducing task success or lengthening its trajectory?
Our primary evidence is a pre-specified, hash-frozen, paired campaign of 2,908 provider-billed Claude Code runs, of which 2,848 were analyzed, covering 103 tasks, seven repositories, and three models. The campaign compared a baseline with two generations of hook-based compression and an API-boundary proxy, within a broader measured program of roughly 5,500 billed executions.
Three findings emerge. First, prompt-cache traffic dominated cost composition. Cache creation and reads accounted for about 87% of reconstructed four-component cost, or about 80% of the actual bill, with an 8.7% dollar-weighted residual that retained telemetry could not attribute. On Haiku 4.5, this residual scaled with thinking effort.
Second, tool-output reduction did not reliably predict billed-cost reduction. An arm that removed 38% of estimated raw tool-output tokens had 6.8% higher paired cost (95% CI: +2.8% to +11.3%), while per-task reduction was only weakly associated with cost change (Pearson r = 0.15, CI crossing zero).
Third, compression can harm task completion by removing action-critical evidence. In a small single-shot study on SWE-bench-derived Go tasks, compression reduced patch application from 27/40 to 15/40 by corrupting verbatim edit anchors, and the compressed grounded arm produced fewer solves at higher observed cost per solve.
We propose a layered evidence standard centered on success-adjusted billed cost rather than token reduction alone.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2607.12161 [cs.CL] |
| (or arXiv:2607.12161v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.12161 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sarel Weinberger [view email]
[v1]
Mon, 13 Jul 2026 21:10:22 UTC (177 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.