# Study finds AI agents waste 18% of tokens on unnecessary tasks

Digest AI · Agents & Tools · published 2026-11-06T15:14:42Z

Canonical: https://digestai.news/story/study-finds-ai-agents-waste-18-of-tokens-on-unnecessary-tasks

## Summary

A new US academic study quantifies how AI agents waste tokens and money during agentic workflows. Researchers tracked token usage across six models—Kimi-K3, Gemini 3.1 Pro, GPT-5.6 Terra, GPT-6 Astra, Claude Sonnet 5, and Claude Fable 5.1—using a ‘shadow’ methodology to measure waste in closed-source systems. They found that **18% of output tokens** were wasted, with **45% of tasks** seeing one agent burn more tokens than another needed to complete the same job successfully. The study identified four waste patterns: **restating known info, duplicate tool requests, padding irrelevant text, and pursuing unproductive detours**. For example, Kimi-K3 wasted **11–24.9%** of tokens on answers and tools, rising to **24.9%** during internal reasoning, while Claude Sonnet 5 performed worst overall. The researchers also found that **41–46% of unnecessary reasoning** could be removed without hurting performance, suggesting significant cost savings are possible. Some waste stemmed from harness flaws, like OpenCode and Kimi Code CLI repeatedly launching tasks in the wrong Python environment, causing retries and higher costs. The study proposes improving model training, reasoning strategies, and harness design to reduce waste, arguing that current AI agents operate like ‘cars that accelerate to max speed on every trip, even when unnecessary.’

## Key points

- Researchers found AI agents waste **18% of output tokens** on unnecessary tasks, with **45% of tasks** seeing one agent burn more tokens than needed
- Removing **41–46% of redundant reasoning** had little impact on performance, suggesting major cost savings are possible without sacrificing results
- Waste patterns include **restating info, duplicate tool calls, padding text, and unproductive detours**, with harness flaws like wrong Python environments causing retries

## Why it matters

The findings expose inefficiencies in AI agent workflows that directly increase costs for users and providers alike. The study also highlights gaps in transparency, as providers like OpenAI and Anthropic have already raised prices while users struggle to understand or control token usage.

## Sources

1. [The AI Detours That You Shouldn’t Be Paying For](https://unite.ai/the-ai-detours-that-you-shouldnt-be-paying-for) (Unite.AI, 2026-11-06)

Part of the developing story: [The Race to Optimize AI Coding Costs](https://digestai.news/thread/bito-releases-governor-to-cut-coding-agent-costs-by-up-to-70) (4 stories)

## Cite

Digest AI, "Study finds AI agents waste 18% of tokens on unnecessary tasks", 6 November 2026, https://digestai.news/story/study-finds-ai-agents-waste-18-of-tokens-on-unnecessary-tasks

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/study-finds-ai-agents-waste-18-of-tokens-on-unnecessary-tasks.json