Anthropic Quietly Cuts Claude Code Weekly Limits by 17% in September

Anthropic will replace Claude Code’s temporary 50% weekly usage boost with a permanent 25% increase on September 14, 2026. The change leaves users with 16.7% less weekly capacity than they have during the promotion, while preserving a 25% increase over the original limits.

The distinction matters for developers who have planned agent runs, automation, and daily coding around the promotional ceiling. Anthropic’s initial announcement emphasized the permanent increase without clearly stating the reduction from current capacity, prompting criticism and a revised explanation.

A higher baseline, a lower ceiling

Anthropic has applied a temporary 50% boost to weekly limits since May, extending the promotion several times. That boost expires on September 13, one day before the permanent limits take effect.

The arithmetic uses two valid reference points. Moving from the original 100% allowance to 125% produces the advertised 25% increase. Moving from the promotional 150% allowance to 125% produces a 16.7% reduction, which Anthropic rounds to 17%.

What changes on September 14

According to Anthropic’s support notice, the revised policy affects weekly usage rather than every Claude Code limit:

• Pro, Max, Team, and seat-based Enterprise plans receive permanent weekly limits set at 125% of the original baseline.
• Shorter rolling usage limits, including the five-hour limits that many users encounter first, remain unchanged.
• The separate weekly limit for Claude Opus disappears.
• Opus usage draws from the shared Claude Code weekly allowance, subject to the remaining session and plan limits.

Removing the dedicated Opus cap gives developers more control over model allocation. A user can direct the full weekly pool toward Opus when a workload benefits from it, rather than preserving a separate Opus budget alongside the general allowance.

Why the announcement drew criticism

Anthropic’s original wording presented the policy as a permanent 25% increase without clearly comparing it with the active 50% promotion. After users challenged that framing, the company deleted or revised the announcement and acknowledged the immediate reduction.

“Compared to today, this works out to a 17% reduction in weekly limits on Claude Code.”

Anthropic also said it is developing better usage-visibility tools. Clearer reporting would help users connect CLI activity, model choice, session limits, and weekly consumption before a cap interrupts work.

Compute sets the limit

Anthropic attributes the change to the cost and availability of computing capacity. Frontier models require substantial accelerator time, and sustained coding-agent workloads can consume more inference capacity than occasional chat sessions.

The promotion was always described as temporary, but repeated extensions encouraged teams to treat the higher ceiling as their working capacity. Reducing that allowance can affect scheduled agents, long-running refactors, continuous integration tasks, and other workflows that consume usage throughout the week.

Plan for the smaller pool

1. Run /usage in the Claude Code CLI and record a representative week under the promotional cap.
2. Treat the new weekly allowance as five-sixths, or 83.3%, of the current allowance.
3. Multiply your current utilization percentage by 1.2 to estimate utilization under the new cap. Using 80% today would consume about 96% after the change.
4. Review scheduled agents and automation on Team or seat-based Enterprise plans for workloads that may exceed the lower weekly ceiling.
5. Revisit model-routing rules if they were designed primarily to preserve the separate Opus allowance.

Developers who stayed below roughly 83% of the promotional cap should fit within the new weekly allowance if their usage remains steady. Teams operating closer to the current ceiling may need to reduce runs, shift work across billing periods, or adjust their plan when the permanent limits begin.