Claude Usage Limit Changes Spark Outrage: Max Account Weekly Quotas Drop Sharply

Claude Max users report 26% weekly quota consumed in 12 hours, coverage expanded to all models, sparking community concern.
Recently, Reddit users reported that the weekly usage quota of Claude Max accounts was abnormally consumed by 26% within 12 hours, far exceeding expectations based on the 5-hour session mechanism. The user noted that the weekly quota's coverage has expanded from specific models to all models, causing usage across different models to count toward a unified quota. Multiple users in the community reported similar issues, speculating that Anthropic may have adjusted the quota calculation method to control costs or that a system bug exists. No official response yet. The article suggests users monitor usage, adjust strategies, and follow official updates.
From "5-Hour Limit" to "Weekly Quota Panic": What Happened to Claude Users?
Recently, a Reddit discussion about Claude usage limits has drawn widespread attention. A Max account user from a startup project posted that their weekly usage quota was consumed by 26% in just 12 hours, and the situation is far worse than before.

According to the user, when they took the screenshot at 12:50 PM on Tuesday, only 12 hours had passed since the last reset. Under Claude's original 5-hour session mechanism, this roughly corresponds to 2.5 session cycles. They used about 90% of one session's quota and about 50% of another, and are currently in the third session. However, their weekly quota has already been consumed by 26%.
What confuses them even more is that last week they also had multiple sessions exceeding limits, but never experienced this kind of rapid weekly quota consumption. Moreover, they are using a Max account provided by the startup project, and the same issue has appeared on other accounts.
Claude's usage limit system typically includes two tiers: short-term session limits (e.g., reset every 5 hours) and long-term period quotas (e.g., weekly or monthly totals). Short-term limits are designed to prevent sudden abuse, while long-term quotas are used to control overall resource consumption and costs. Different subscription tiers (such as Pro and Max) may have different quota thresholds and coverage, and the platform reserves the right to adjust policies without prior notice. Understanding this two-tier structure helps analyze the "abnormal weekly quota consumption" reported by users—that is, short-term sessions are not over limit, but the long-term quota drops rapidly.
Key Change Observed by Users: Expanded Coverage of Weekly Quota
The user pointed out a key detail: until last week (i.e., 12 hours ago), the weekly quota only applied to the "fable model" (presumably a specific model, possibly a typo or alias). Now, the weekly quota seems to apply to all models.
This means that when users use different models, the consumed quota may all be counted uniformly toward the weekly quota. For users who rely on Claude for multiple tasks, this undoubtedly increases the difficulty of quota management.
The user also mentioned that Claude previously removed the 5-hour limit for some users, but this is not the core of the current problem. The real issue is the abnormally accelerated consumption of the weekly quota.
In AI services, model routing refers to assigning requests to models of different sizes or costs based on task type (e.g., lightweight models handle simple queries, large models handle complex reasoning). If the weekly quota previously only applied to specific models (such as the "fable model" mentioned in the article), users could bypass the quota by switching models; but when the quota covers all models, usage of any model consumes the unified weekly quota. This change may reflect the platform extending cost control from a single model to the entire model pool, thereby affecting users' quota management strategies.
Community Feedback: Not an Isolated Case
The post explicitly states that the same problem also appears on other accounts. This suggests that the phenomenon may not be an isolated incident but a platform-level adjustment or malfunction.
In the Reddit discussion, many users expressed confusion about Anthropic's usage limit strategy. Some users speculated that Anthropic may have adjusted the quota calculation method to control costs, while others believed it could be a system bug causing abnormal quota consumption.
Interestingly, as of the time of posting, Anthropic had not provided a clear explanation. Users can only speculate about possible causes through community exchanges.
Impact on Users and Suggested Responses
For users who rely heavily on Claude, the rapid consumption of the weekly quota means they may hit the upper limit earlier, thereby affecting their workflows. Especially for startup teams using Max accounts, quota limits directly affect project progress efficiency.
At present, users can take the following measures:
- Monitor usage: Regularly check account usage statistics to understand consumption across models.
- Adjust usage strategy: If the weekly quota covers all models, prioritize allocating quota to critical tasks.
- Follow official updates: Pay attention to Anthropic's announcements or updates to get the latest information on quota policies.
- Community exchange: Share experiences on platforms like Reddit to learn whether other users are encountering similar issues.
Summary
The sudden change in Claude usage limits has made some users uneasy. The weekly quota has expanded from specific models to all models, and consumption speed is abnormal, possibly due to platform policy adjustments or technical failures. Before official clarification, users need to closely monitor usage and flexibly adjust their usage patterns.
We will continue to follow this matter and bring the latest developments promptly.
Related articles

Running a Local LLM on a Laptop CPU: Ollama + Qwen3 Gives a Desktop Pet a Brain
An ordinary laptop without a GPU can run a local LLM: a Bilibili creator uses Ollama plus open-source Qwen3 8B on pure CPU to give an original desktop pet a local brain, with DeepSeekHarness handling plugin development and three rounds of debugging — 0 yuan, works offline, privacy stays on-device.

AgentScope 2.0 in Practice: A Guide to Enterprise-Level AI Agent Development in Java
AgentScope 2.0 is Alibaba's Java enterprise-grade distributed agent framework. This article explains why it suits server-side AI applications, contrasts client-side AI apps in permissions, state management, and concurrency, and outlines a practical learning path from environment setup to project integration.

GPT-6 Astra Directly Edits Weights to Fix Models? Full Analysis of the MIT Paper and Leaks
MIT paper shows GPT-6 Astra repairs small models by directly editing weights, raising math scores from 7.9% to 48.1%; OpenAI internal charts hint an unreleased model leads on BenchCAD 2.0; Fable 5.5 pure-code movies and the Claude WiFi radar incident analyzed.