تخطى إلى المحتوى الرئيسي

Claude Code Limits: 5-Hour Caps Doubled

Rui Dai
Rui Dai Engineer
مشاركة

You sent thirty prompts before lunch. Two months ago, that was half your daily Claude Code budget on Pro — and during peak hours, it was even less. As of May 2026, the math has changed.

What Changed on May 6

On May 6, 2026, Anthropic shipped three changes at once. The 5-hour rate limits on Claude Code doubled for every paid plan. The peak-hour throttle — which during busy windows quietly cut available quota on Pro and Max by up to half — was removed entirely. And API rate limits for Opus models went up substantially across the board.

The practical effect: three days of typical Claude Code usage now costs you what one day did in April. For lighter sessions, you may never see the wall.

The Weekly Bump (and the July 13 Expiration)

On May 13, one week after the 5-hour change, Anthropic followed up with a 50% increase to weekly limits across Pro, Max, Team, and seat-based Enterprise plans. Unlike the rate-limit change, this one has an expiration date: July 13, 2026, unless extended.

The shape of usage matters here. The 5-hour window is a rolling rate limit — your usage over the last 300 minutes. The weekly cap is the harder ceiling, the one heavy users actually crash into. The May 6 change widened your cushion against short-burst rate limiting. The May 13 change widened the ceiling itself, for two months.

Claude Code May 2026 usage limit timeline

The weekly bump arrived on May 13 and is scheduled to expire on July 13, 2026 unless Anthropic extends it.

Plan by Plan

What you get depends on which tier you pay for. The free plan was excluded from both bumps.

  • Pro ($20/mo): doubled 5-hour cap — roughly 45 prompts per window becomes roughly 90, with the usual caveats about message length and model. No more peak-hour penalty. Weekly cap is 50% higher through July 13.
  • Max 5x ($100/mo): doubled 5-hour cap, no peak penalty, weekly Opus 4.7 budget moves from around 50 hours to around 75.
  • Max 20x ($200/mo): doubled 5-hour cap, no peak penalty, weekly Opus 4.7 budget moves from around 200 hours to around 300.
  • Team and seat-based Enterprise: both bumps apply per seat. Admin-side billing math is unchanged.
  • Free: unchanged. Limits sit where they did in April.

Why Now

The capacity came from somewhere. Anthropic signed a compute deal with SpaceX for Colossus 1 — more than 300 megawatts of new capacity and 220,000-plus NVIDIA GPUs coming online inside the month.

That's the supply side. The demand-side pressure is Codex. OpenAI's coding agent has been gaining ground on long-running tasks, and Claude Code's reputation for hitting walls during heavy sessions had become a recurring complaint. Doubling the 5-hour window and killing peak throttling is the most direct answer Anthropic could ship: more headroom, no price change.

What This Means in Practice

If you've been pacing yourself — closing sessions early, batching tool calls, dropping to Haiku for routine edits — you can stop. Or at least stop reflexively. Three workflow shifts make sense after these changes.

First, try Opus 4.7 for tasks you previously routed to Sonnet. The Opus API rate limits went up the same day, so even tight tool-call loops are less likely to throttle.

Second, stop scheduling around peak hours. The penalty is gone on Pro and Max, so 2pm Pacific is identical to 2am Pacific for quota purposes. You no longer pay a tax for working during normal hours.

Third, don't restructure your workflow assuming the weekly +50% sticks. It expires July 13, 2026. If your day-to-day depends on the new ceiling, plan for the reset — or watch for an extension announcement closer to that date.

The May 15 Counter Reset

Two days after the weekly bump, on May 15, Anthropic flushed every user's 5-hour and weekly counters back to zero — a one-time release valve to let people actually feel the new ceilings. That reset is over; counters tick normally again. If you tested the new limits that week and burned a lot of quota, you may have noticed they topped up faster than expected. That was why.

The Bigger Picture

Two months ago, the recurring Claude Code question was 'how do I avoid hitting the limit?' After May, it's closer to 'how much can I ship before July 13?' That's a healthier place to be — for now. Whether the weekly bump becomes permanent is the next thing to watch.

Rui Dai
كتبهRui Dai Engineer

Hey there! I’m an engineer with experience testing, researching, and evaluating AI tools. I design experiments to assess AI model performance, benchmark large language models, and analyze multi-agent systems in real-world workflows. I’m skilled at capturing first-hand AI insights and applying them through hands-on research and experimentation, dedicated to exploring practical applications of cutting-edge AI.