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

arXiv: https://arxiv.org/abs/2604.22750 · HTML: https://mlsanity.com/paper/2604.22750/

## Key facts

- Cited by 33 papers in the map (data through 14 Oct 2026).
- Placed in Agent memory systems, one of 16 directions in LLM agents and tool use.
- Builds on 18 of its 22 references that are in the map.
- First posted on arXiv 24 Apr 2026.

## Placed in

- LLM agents and tool use / [Agent memory systems](https://mlsanity.com/agents/agent-memory-systems/)

## Builds on (most cited first, 18 of 18)

- [Self-Refine: Iterative Refinement with Self-Feedback](https://mlsanity.com/paper/2303.17651/) (2023)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Voyager: An Open-Ended Embodied Agent with Large Language Models](https://mlsanity.com/paper/2305.16291/) (2023)
- [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://mlsanity.com/paper/2408.03314/) (2024)
- [Language Models (Mostly) Know What They Know](https://mlsanity.com/paper/2207.05221/) (2022)
- [OpenHands: An Open Platform for AI Software Developers as Generalist Agents](https://mlsanity.com/paper/2407.16741/) (2024)
- [Executable Code Actions Elicit Better LLM Agents](https://mlsanity.com/paper/2402.01030/) (2024)
- [RepoBench: Benchmarking Repository-Level Code Auto-Completion Systems](https://mlsanity.com/paper/2306.03091/) (2023)
- [CRUXEval: A Benchmark for Code Reasoning, Understanding and Execution](https://mlsanity.com/paper/2401.03065/) (2024)
- [Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning](https://mlsanity.com/paper/2502.18080/) (2025)
- [Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?](https://mlsanity.com/paper/2502.12215/) (2025)
- [Inverse Scaling in Test-Time Compute](https://mlsanity.com/paper/2507.14417/) (2025)
- [Budget-Aware Tool Use Enables Effective Agent Scaling](https://mlsanity.com/paper/2511.17006/) (2025)
- [OptimalThinkingBench: Evaluating Over and Underthinking in LLMs](https://mlsanity.com/paper/2508.13141/) (2025)
- [Efficient Agents: Building Effective Agents While Reducing Cost](https://mlsanity.com/paper/2508.02694/) (2025)
- [SWE-Effi: Re-Evaluating Software AI Agent System Effectiveness Under Resource Constraints](https://mlsanity.com/paper/2509.09853/) (2025)
- [Tokenomics: Quantifying Where Tokens Are Used in Agentic Software Engineering](https://mlsanity.com/paper/2601.14470/) (2026)
- [The Price Reversal Phenomenon: When Cheaper Reasoning Models Cost More](https://mlsanity.com/paper/2603.23971/) (2026)

## Built on by (most cited first, 30 of 33)

- [SWE-MeM: Learning Adaptive Memory Management for Long-Horizon Coding Agents](https://mlsanity.com/paper/2606.28434/) (2026)
- [BAGEN: Are LLM Agents Budget-Aware?](https://mlsanity.com/paper/2606.00198/) (2026)
- [Token Economics for LLM Agents: A Dual-View Study from Computing and Economics](https://mlsanity.com/paper/2605.09104/) (2026)
- [CoACT: Action-Preserving Observation Compression for Coding Agents](https://mlsanity.com/paper/2607.02911/) (2026)
- [FrogNano: Training a 4B Coding Agent via Online Task Synthesis](https://mlsanity.com/paper/2609.07925/) (2026)
- [When Agents Coordinate: Measuring Coordination in Multi-Agent AI Coding](https://mlsanity.com/paper/2608.16801/) (2026)
- [Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training](https://mlsanity.com/paper/2605.28008/) (2026)
- [What Should Agents Say? Action-state Communication for Efficient Multi-Agent Systems](https://mlsanity.com/paper/2606.05304/) (2026)
- [Reasoning effort, not tool access, buys first-try reliability in agentic code generation: an observational study](https://mlsanity.com/paper/2607.02436/) (2026)
- [SMetric: Rethink LLM Scheduling for Serving Agents with Balanced Session-centric Scheduling](https://mlsanity.com/paper/2607.08565/) (2026)
- [Markets, Not Planners: Decentralized Orchestration of LLM Agents with Private Information](https://mlsanity.com/paper/2608.23867/) (2026)
- [Cost-Utility Alignment in LLM Agent Trajectories:Profiling,Attribution,Diagnosis,Adaptation,and Evaluation](https://mlsanity.com/paper/2608.26195/) (2026)
- [Does Code Cleanliness Affect Coding Agents? A Controlled Minimal-Pair Study](https://mlsanity.com/paper/2605.20049/) (2026)
- [The Token Tax of Epistemic Accuracy: Comparing RAG and Long-Context Architectures for Document-Grounded Generative AI Applications](https://mlsanity.com/paper/2606.20898/) (2026)
- [The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming Languages](https://mlsanity.com/paper/2607.22807/) (2026)
- [Can your AI agent be cheaper? Investigating the effects of task specifications on token spend in agentic coding tasks](https://mlsanity.com/paper/2608.25399/) (2026)
- [MEMOA: Massive Mixtures of Online Agents via Mean-Field Decentralized Nash Equilibria](https://mlsanity.com/paper/2605.05492/) (2026)
- [AI Tokenomics: The Economics of Tokens, Computation, and Pricing in Foundation Models](https://mlsanity.com/paper/2606.24616/) (2026)
- [ACEM: A Cost Estimation Model for Agentic Software Engineering](https://mlsanity.com/paper/2608.02582/) (2026)
- [Predicting Task Difficulty Without Rollouts](https://mlsanity.com/paper/2608.05797/) (2026)
- [Can a Lightweight Multimodal Model Estimate LLM Reasoning Performance? A Study for Compute-Optimal Document Inference](https://mlsanity.com/paper/2608.18591/) (2026)
- [AgentDecarbonizer: Carbon-Aware Execution for AI Agents](https://mlsanity.com/paper/2608.20566/) (2026)
- [LLM Agents as Computational Typologists](https://mlsanity.com/paper/2609.07791/) (2026)
- [AgentKV: Phase-Aware KV Eviction for Agentic LLMs](https://mlsanity.com/paper/2609.14872/) (2026)
- [Analyzing and Mitigating Cost-Inefficient Behaviors in Coding Agents](https://mlsanity.com/paper/2609.30725/) (2026)
- [ResonAct: Streaming Metrics for Runtime Diagnosis and Self-Healing in Multi-Agent Systems](https://mlsanity.com/paper/2609.34701/) (2026)
- [TokenCast: Forecasting Token Consumption During LLM Agent Execution](https://mlsanity.com/paper/2609.35760/) (2026)
- [LongHarness Bench: Stress-Testing Language Model Harnesses for Long-Context Reasoning](https://mlsanity.com/paper/2609.38137/) (2026)
- [Complex Agents, Shallow Tests: Demystifying and Enhancing Test Adequacy of Agent Harness in the Wild](https://mlsanity.com/paper/2610.04921/) (2026)
- [Disentangling Task Difficulty from Run-Level Failure in Agent Failure Prediction](https://mlsanity.com/paper/2610.05572/) (2026)

## Abstract (arXiv)

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.