All of that has been conquered and Claude is starting to grow thin.
Earlier this week, Anthropic adjusted usage limits for Claude Free, Pro, and Max subscribers. The weekly limit will not change, but users will approach the cap faster if they use Claude during peak hours (which Anthropic defines as 5 a.m. to 11 a.m. Pacific Time).
“We’ve achieved a number of efficiencies to offset this, but around 7% of our users will hit session limits that weren’t there before, especially in the pro tier,” Thariq Shihipar, who works on Claude, wrote about X. “If you’re running a token-intensive background job, shifting it to off-peak hours will stretch your session limits further.”
Adjust the free/pro/maximum subscription 5-hour session limit during peak periods to address increased demand for Claude. Weekly limits remain unchanged.
On weekdays from 5am to 11am PT/1pm to 7pm GMT, you’ll be able to get past the 5-hour session limit faster than before.
— Tariq (@trq212) March 26, 2026
“I know this is frustrating,” Schichper wrote. “We continue to invest in efficient expansion and will keep you informed of our progress.”
Mainstream interest in Anthropic has increased since CEO Dario Amodei refused to give the Department of Defense unfettered access to the company’s AI models. Amodei previously said the company’s core focus is on its enterprise business.
Like its competitors, Anthropic is experiencing growing pains as it tries to balance the available compute. Tools like OpenClaw, an open source AI agent, allow users to harness the full potential of AI models like never before. At the same time, the explosion of agents means users require more computing power than ever before.
OpenAI announced earlier this week that it would be retiring Sora, its once-popular TikTok-like AI video generation app, to refocus computing on its core services.
In December, Amodei appeared to ridicule OpenAI CEO Sam Altman’s ambitious spending plans for mega data centers to meet future computing demands. Manufacturers of AI frontier models and big techs like Microsoft and Google have committed enormous resources to staying ahead of the AI race.
“I think there’s some unmitigated risk here, and I definitely don’t want to deny that,” Amodei told The New York Times’ Dealbook Summit. “But at the same time, I think there are some players who don’t manage that risk very well and are taking unwise risks.”