Anthropic announced it will permanently raise Claude Code's standard weekly usage limits by 25% on September 14 for Pro, Max, Team, and seat-based Enterprise plans. The framing is generous. The reality is a cut.
The math behind the announcement
Claude Code currently runs on a temporary 50% increase to weekly limits, which Anthropic put in place earlier and extended through September 13. If the original weekly allowance was 100, the temporary boost gives you 150 today. On September 14, that drops to 125.
The 25% increase is measured against the original baseline, not against the current temporary boost. Compared to what you have right now, the new permanent limit is a 17% reduction. Anthropic's original announcement on X framed it as a permanent increase. The company later deleted that thread and posted a clarification that acknowledged the reduction directly: "Compared to today, this works out to a 17% reduction in weekly limits on Claude Code."
What the limits actually control
Weekly limits in Claude Code do not translate into a fixed number of prompts. Anthropic says usage varies based on conversation length, model choice, tool usage, and effort level. A short conversation with Haiku uses fewer tokens than a long session with Opus that invokes tools repeatedly. The limit is measured in tokens consumed, not in turns taken, which makes the exact impact hard to calculate without knowing your usage pattern.
The temporary 50% boost remains available through September 13. The new permanent limits take effect on September 14.
Why this matters for developers using Claude Code
Claude Code has become a primary coding tool for many developers since its launch. Teams that built workflows around the current limit will need to adjust. The 17% reduction is not catastrophic, but it is a reduction in a tool that many developers depend on for daily work. The announcement also hints at further changes: Anthropic said it is working on changes that will "make it feel like you're getting more from Claude, while having more visibility and control of your usage." That language suggests the limit structure itself may change, not just the numbers.
For teams evaluating AI coding tools, the takeaway is that usage limits on cloud-hosted AI services are not fixed contracts. They can be adjusted by the provider at any time, and the framing of those adjustments does not always match the impact. A 25% increase sounds like a win until you check what it is 25% of.