What did Anthropic announce about Claude Code’s weekly usage limits for Pro, Max, Team, and seat-based Enterprise subscribers beginning September 14, how does the change compare mathematically with the temporary 50% boost available since May 13 and result in a 17% reduction from

Anthropic said that from September 14 it would permanently set Claude Code’s standard weekly limits at 25% above the pre-promotion baseline for Pro, Max, Team, and legacy seat-based Enterprise subscribers. That is a permanent improvement over the old baseline, but a 16.7% decrease versus the temporary capacity users had immediately before the change. 12

• The math: Let the former standard allowance be 100 units. The May 13 promotion made it 150; September’s permanent limit is 125. Thus users lose 25 of their current 150 units: 25 / 150 = 16.7%, generally reported as a 17% reduction. The 50% promotion had been extended several times. 23

• Messaging and reaction: The initial framing—“permanently raising” limits by 25%—prompted criticism because it compared with an older baseline rather than the capacity customers were actually receiving. A Community Note highlighted the 17% effective reduction. Users characterized the wording as misleading or a bait-and-switch; Anthropic’s position was that the 50% amount was explicitly a temporary promotion, while the 25% uplift is permanent. 12

• What the allowance does—and does not—mean: It is not a guaranteed number of prompts or coding hours. Real consumption depends on the model selected, prompt and output-token volume, context accumulated in a session, tool/agent activity, task complexity, and service-demand controls such as peak-hour throttling. Anthropic has previously used peak-hour controls under which usage can be consumed faster during high demand. 4

• Future quota controls: Anthropic indicated it was working toward clearer usage visibility and more flexible ways for heavy users to manage or obtain capacity. However, the retrieved evidence does not establish final specifications, pricing, or release dates for a usage dashboard, overages, or add-on quota; those should be treated as previews rather than committed product terms.

• Who is unaffected: The announcement applies to the specified paid Claude Code subscriptions—notably including only legacy seat-based Enterprise users. Free-plan users and customers using Claude through API/consumption-based arrangements are outside that stated weekly-promotion scope; Enterprise customers on other billing structures should confirm their contract terms. 3

Industry context

This fits a broader move from simple “generous subscription” allowances toward explicit capacity management and metering for expensive AI workloads.

• OpenAI restored five-hour caps for ChatGPT Plus users’ Codex and ChatGPT Work access on August 25 after a period governed only by a weekly limit, citing the need to manage compute demand; Pro, Enterprise, and Edu were reported as unaffected. 56
• GitHub moved Copilot to usage-based billing on June 1, replacing request-based billing for most customers; its enterprise offerings now include defined AI-credit allocations, making higher-cost usage more visibly metered. 78
• Google shifted Gemini toward “compute-specific” limits rather than fixed daily feature-generation counts, with limits refreshing every five hours; Gemini Notebook was slated to adopt the approach in September. 9

So the accurate interpretation is mixed: Anthropic is permanently raising the historic baseline, but it is also withdrawing part of a long-running temporary benefit. The common industry direction is not necessarily a uniform price increase; it is tighter linkage between access, workload cost, model choice, and available compute.