BotBeat
...
← Back

> ▌

AnthropicAnthropic
UPDATEAnthropic2026-04-02

Anthropic Addresses Unexpected Token Consumption Issues in Claude Code

Key Takeaways

  • ▸Claude Code users report token limits being exceeded much faster than expected across multiple subscription tiers
  • ▸Some users experienced complete daily budget depletion in minutes, particularly during looped coding operations
  • ▸Anthropic has prioritized fixing the issue but has not provided a timeline or detailed explanation of the root cause
Source:
Hacker Newshttps://www.bbc.com/news/articles/ce8l2q5yq51o↗

Summary

Anthropic is investigating a significant issue where Claude Code users are hitting their token usage limits far faster than anticipated, with the company calling it a "top priority" for resolution. Users across free and paid tiers—ranging from $20 to $200 per month—have reported dramatically accelerated token depletion, with some experiencing budget exhaustion within minutes during looped coding sessions or even from single-sentence replies. The problem has become acute following Anthropic's recent introduction of peak-hour throttling, which consumes tokens more quickly during high-demand periods, exacerbating the opacity around token consumption rates for specific tasks. The issue threatens to disrupt developer workflows that depend on Claude Code as part of their daily toolkit.

  • Recent peak-hour throttling implementation may be contributing to the accelerated token consumption problem
  • Token consumption mechanics remain opaque to users, making it difficult to predict costs for specific tasks
Large Language Models (LLMs)Generative AIAI AgentsMarket TrendsProduct Launch

More from Anthropic

AnthropicAnthropic
RESEARCH

Anthropic Study Reveals AI Agent Memory Retrieval Accuracy at Just 9%, Exposing Infrastructure Challenges

2026-07-04
AnthropicAnthropic
POLICY & REGULATION

Anthropic Receives Cease and Desist Over Claude Desktop Privacy Violations

2026-07-04
AnthropicAnthropic
RESEARCH

Research: How URLs in Prompts Can Influence LLM Outputs Toward Training Data

2026-07-03

Comments

Suggested

MicrosoftMicrosoft
RESEARCH

Microsoft's Leaked 'Aion' Project Reveals Vision for Copilot-First Operating System

2026-07-04
Google / AlphabetGoogle / Alphabet
RESEARCH

Stanford Researchers Use Multi-Agent AI and Reinforcement Learning to Improve HIP Kernel Generation for AMD GPUs

2026-07-04
Rampart (Independent Project)Rampart (Independent Project)
INDUSTRY REPORT

First Large-Scale Study Shows AI Adoption Drives Job Growth, Not Displacement

2026-07-04
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us