A developer paying for Anthropic's premium Claude AI service reported hitting usage limits in just 90 minutes during routine work, raising questions about the platform's quota system and pricing transparency.
The Problem: Max 5x Plan Exhausted Quickly
GitHub user molu0219, who subscribes to Claude's "Max 5x" plan at $100 per month, posted a bug report titled "[BUG] Pro Max 5x Quota Exhausted in 1.5 Hours Despite Moderate Usage" on the platform's issue tracker.
Claude operates on a 5-hour rolling window for usage limits. When users reach their quota within that period, they're locked out until the full 5 hours elapse. The "Max 5x" plan offers five times the usage allowance of the standard $20 "Pro" plan, but molu0219 found even this wasn't sufficient.
What Happened During Testing
On the day in question, molu0219 conducted intensive development work from 3 PM to 8 PM, generating 2,715 API calls with maximum context length reaching approximately 970,000 tokens. The system's automatic context summarization feature triggered twice during this session. While hitting limits during this heavy usage was understandable, what followed wasn't.
After 8 PM, during what molu0219 describes as "light development work and Q&A"—typical usage patterns—the quota was exhausted in just 90 minutes.
The Suspected Culprit: Cached Sessions
Investigation revealed that background Claude sessions, left open but inactive, were performing extensive cache reads. While cached inputs are billed at one-tenth the cost of regular inputs, molu0219 suspects they're being counted at full volume toward usage limits. This creates a problematic situation where Claude's advertised 1-million-token context window—meant to handle large amounts of information at once—might actually work against users.
"If cache reads are counted at full rate," molu0219 wrote, "then having a larger context window increases input tokens per API call, making it easier to hit usage limits." The developer also noted that "idle sessions in the background that users aren't actively using shouldn't consume massive API resources."
Anthropic's Response and Industry Context
A Claude Code team member named Boris responded in the comments, suggesting potential fixes like "reducing the standard context window" and "actively working on background task cleanup."
Anthropic finds itself in a challenging position. The company has seen revenue triple in just three months and announced plans to expand computing capacity, but immediate resources appear strained. In response to mounting demand, Anthropic has tightened usage limits, prompting user complaints about degraded performance and increased consumption due to shorter cache expiration times.
The situation highlights growing pains in the AI industry as services like Claude balance rapid growth with sustainable resource management and transparent pricing models.