How Do AI Agents Spend Your Money? Analyzing and Predicting Token Consumption in Agentic Coding Tasks

2026-04-24

AI Summary

This study addresses the unclear usage patterns and efficiency of large language model (LLM) tokens in AI agents performing complex tasks. By systematically analyzing coding trajectories of eight state-of-the-art models on SWE-bench Verified, the work reveals that agent-based tasks can consume up to 1,000 times more tokens than conventional code generation tasks, with execution-to-execution variability reaching 30-fold within the same task. Input tokens dominate computational cost, while human-assessed task difficulty shows only weak correlation with actual token expenditure. Moreover, models consistently and substantially underestimate their own token usage, with prediction accuracy yielding a maximum correlation coefficient of merely 0.39. Notably, Kimi-K2 and Claude-Sonnet-4.5 consume, on average, 1.5 million more tokens than GPT-5. Through empirical trajectory analysis, statistical modeling, and cross-model comparison, this research establishes a foundational understanding for optimizing AI agent resource efficiency.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

The wide adoption of AI agents in complex human workflows is driving rapid growth in LLM token consumption. When agents are deployed on tasks that require a significant amount of tokens, three questions naturally arise: (1) Where do AI agents spend the tokens? (2) Which models are more token-efficient? and (3) Can agents predict their token usage before task execution? In this paper, we present the first systematic study of token consumption patterns in agentic coding tasks. We analyze trajectories from eight frontier LLMs on SWE-bench Verified and evaluate models' ability to predict their own token costs before task execution. We find that: (1) agentic tasks are uniquely expensive, consuming 1000x more tokens than code reasoning and code chat, with input tokens rather than output tokens driving the overall cost; (2) token usage is highly variable and inherently stochastic: runs on the same task can differ by up to 30x in total tokens, and higher token usage does not translate into higher accuracy; instead, accuracy often peaks at intermediate cost and saturates at higher costs; (3) models vary substantially in token efficiency: on the same tasks, Kimi-K2 and Claude-Sonnet-4.5, on average, consume over 1.5 million more tokens than GPT-5; (4) task difficulty rated by human experts only weakly aligns with actual token costs, revealing a fundamental gap between human-perceived complexity and the computational effort agents actually expend; and (5) frontier models fail to accurately predict their own token usage (with weak-to-moderate correlations, up to 0.39) and systematically underestimate real token costs. Our study offers new insights into the economics of AI agents and can inspire future research in this direction.

Problem

• token consumption
• AI agents
• agentic coding tasks
• token efficiency
• cost prediction

Innovation

• token consumption
• AI agents
• agentic coding
• token efficiency
• cost prediction