
Anthropic adjusted Claude's session limits so that usage is more aggressively constrained during peak hours. The company said users are likely to burn through their limits faster when demand is high, amid compute pressure.
The change was tied to Claude's rapid growth with paying users. As demand surged, Anthropic faced additional strain on the underlying compute capacity needed to run responses. Rather than allowing unlimited peak utilization, the firm tightened throttling so that service levels could be maintained for more customers.
For enterprise and power users, limit changes directly affect reliability and planning: workflows that depend on steady Claude availability can see slower throughput or earlier cutoff during busy periods. For the broader AI market, the episode underscores a recurring pattern -- popular models often require operational rationing as usage grows faster than the capacity needed to serve everyone.
The coverage also suggests these limit tweaks are part of an ongoing balancing act between growth, user experience, and cost of inference. But the specific engineering details of the capacity constraints weren't included in the provided story, so the most concrete takeaway remains the operational decision: more peak-hour throttling for Claude.