Anthropic Tweaks Timed Usage Limits to Discourage Claude Demand During Peak Hours

Anthropic Tweaks Timed Usage Limits to Discourage Claude Demand During Peak Hours

Anthropic has quietly restructured how Claude Code's five-hour session limits are metered — and if your workflows have been burning through limits faster than usual, this is why. Thariq Shihipar from Anthropic engineering posted on X that the company is now reducing effective token throughput during peak hours (05:00–11:00 PT / 13:00–19:00 GMT) while expanding it during off-peak windows. The net weekly token budget is unchanged, but the same five-hour session now counts tokens faster at peak and slower at off-peak. The Register reports that approximately 7% of users — primarily those running token-intensive background jobs — will hit session limits they didn't previously hit under the old uniform metering.

The practical impact is sharpest for anyone running CI/CD agentic loops, long multi-file refactor sessions, or multi-agent orchestration during US business hours. Those workflows can now exhaust a day's effective budget significantly faster than they did a week ago. The straightforward mitigation is to shift heavy token jobs to off-peak windows — US evenings or overnight — where the same token spend now goes further. API customers paying per token are not affected; only flat-rate subscription holders (Pro, Max, Team) see the change. The Register also called out that Anthropic still does not publicly disclose how many tokens constitute a five-hour session window, making the actual impact difficult to calculate independently. Monitoring daily and weekly progress via the new usage dashboard is currently the only reliable way to gauge exposure.

Read the full article at The Register →