Anthropic Reduces Claude Usage Allowances During Peak Hours to Manage Capacity

Anthropicadjusted its usage limits for Claude subscription customers on Wednesday by compressing session allowances during peak demand periods, aiming to balance demand with infrastructure capacity without reducing overall weekly limits.

According to Thariq Shihipar, a member of Anthropic's technical team, the changes affect the five-hour session limits for Free, Pro ($20/month), Max 5x ($100/month), and Max 20x ($200/month) subscribers. During peak hours—05:00 to 11:00 PT or 13:00 to 19:00 GMT—users will deplete their five-hour session allowance faster than the standard five-hour window. Outside peak hours, the allocation remains linear.

The mechanism relies on Anthropic's opaque token-to-time calculation. The company ties hourly usage limits to token consumption but does not publicly disclose the exact conversion rate. "Your usage is affected by several factors, including the length and complexity of your conversations, the features you use, and which Claude model you're chatting with," according to Anthropic's documentation.

Impact Assessment

Shihipar stated that approximately 7% of users will hit session limits they would not have encountered under the previous scheme, with Pro tier subscribers disproportionately affected. However, he emphasized that weekly usage limits remain unchanged: "Overall weekly limits stay the same, just how they're distributed across the week is changing."

Users running token-intensive background tasks can mitigate the impact by scheduling them during off-peak hours, when Anthropic has expanded capacity. The company simultaneously invested in additional capacity during lower-demand periods to offset peak-hour restrictions.

Customers have access to a dashboard displaying progress toward both daily five-hour session limits and weekly usage allowances. Users who exceed limits face lockout unless they purchase additional usage.

Pricing and Access Models

Anthropicoffers Claude through two channels: an API with published per-token pricing across multiple categories (Base Input Tokens, Cache Write operations, Cache Hits, and Output Tokens), and subscriptions with unpublished usage limits. The subscription model provides predictable monthly costs but opaque consumption ceilings.

Shihipar acknowledged customer frustration with the adjustment but framed it as temporary: "I know this was frustrating. We're continuing to invest in scaling efficiently. I'll keep you posted on progress."

What This Means

Anthropic is prioritizing load balancing over uniform user experience. The peak-hour throttling represents a soft form of demand destruction—making peak-hour usage less attractive without outright blocking it or raising prices. For developers, this creates an incentive to shift compute-heavy workloads to night hours (California time), effectively subsidizing Anthropic's infrastructure investment through user behavior modification. The lack of transparent token metrics means users cannot accurately forecast when they'll hit limits, reducing pricing predictability compared to Anthropic's per-token API model.