Claude Max Used to Die by Wednesday. Anthropic's Fix Expires August 19.

On May 6, 2026, Anthropic doubled Claude Code's five-hour usage limits and removed the peak-hour throttle it had run for the previous two months.

That single announcement answers the complaint this article made in April: that Claude Max, sold as the plan for serious work, was quietly running out by midweek. Four months later, the floor under Max is real and measurably higher than it was then. The headroom keeping Anthropic's heaviest users comfortable right now, though, is not that permanent fix. It is a promotional weekly boost, renewed four times since May, currently scheduled to end on August 19, 2026, two weeks from this article's publication.

What actually changed since April

The original complaint had three parts, and Anthropic has answered two of them directly. First, weekly caps stacked on top of the older rolling five-hour window, introduced in 2025 to stop 24-hour background scripting and account sharing. Second, Anthropic confirmed on March 31, 2026, that it had tightened five-hour limits during weekday peak hours, 5 to 11 a.m. Pacific, and admitted people were hitting Claude Code limits faster than expected. Third, users reported Claude Code hanging, looping, and losing context mid-session during that same stretch.

The peak-hour throttle is gone. Anthropic removed it on May 6, alongside the five-hour doubling, for every Pro, Max, Team, and seat-based Enterprise account. That leaves the weekly cap and the reliability complaints, which is where the rest of this story lives.

Where the extra capacity actually came from

The fix was not a policy memo. It was compute. Anthropic's May 6 announcement coincided with a deal for the full output of SpaceX's Colossus 1 data center in Memphis: more than 220,000 Nvidia GPUs and over 300 megawatts, brought online within the month, according to reporting from The Decoder that cited Anthropic's own statement on the partnership. Anthropic also raised Claude Opus API rate limits by 2 to 16 times in the same window.

That scale of new compute is why the doubling stuck instead of quietly reversing a month later, the way the March tightening did.

The safety valve nobody asked for

Anthropic also shipped something new: pay-as-you-go usage credits for Pro, Max 5x, and Max 20x plans, according to its help center. Once you hit your plan's limit, you can switch to standard API-rate billing rather than waiting for a reset, up to a $2,000 daily cap, with optional monthly spending controls. It is not free capacity. It is a way to buy your way past a wall that used to just stop you.

Pro and Team Standard users on the Fable 5 model got a blunter version of the same idea in July: a one-time $100 transition credit, then metered pricing at $10 per million input tokens and $50 per million output tokens, roughly one long autonomous coding session per credit.

The boost that keeps almost expiring

Here is the part that has not settled. Since mid-May, Anthropic has run a promotion that raises Claude Code's weekly limit by 50%, across the CLI, IDE extensions, desktop app, and web, for Pro, Max, Team, and legacy Enterprise seats.

It was due to expire in June. It did not. It was due to expire July 19. It did not. Anthropic pushed it to August 19 instead, its fourth extension. Max and Team Premium users also kept access to the Fable 5 model at up to 50% of that same weekly pool, past its original July 19 cutoff.

Four extensions is a pattern, not an accident. It also means the number everyone is budgeting around right now, the +50%, has never once been Anthropic's stated permanent policy.

Who this helps, and who it still does not

Anyone who scaled Claude Code into daily use since May is living on a floor that is objectively better than April's: doubled base capacity, no peak-hour penalty, an overage option instead of a hard wall. For a two-person team running one agent each, that is the difference between finishing a sprint and stopping short of it.

It does less for the exact complaint that started this story. Cursor users comparing seat cost still get a plan with published usage tiers rather than a rolling promotion. And nothing here fixes the reliability complaints, the hanging, the looping, the context loss, that users reported alongside the March tightening. Anthropic's compute deal bought headroom. It did not publish a fix for those bug reports.

The floor matters more than the ceiling

The strongest case against treating August 19 as a cliff is that the permanent fix already happened. The May 6 doubling and the removed peak-hour throttle are the baseline now, not a promotion. Losing the extra 50% on Claude Code would return most users to a floor that is still meaningfully better than the one this article described in April, not back to a plan that dies by Wednesday.

That is true for anyone running Claude Code the way they did in April. It is not true for teams that changed their workflow to match the boosted numbers: running a second agent in parallel, keeping longer sessions open, treating headroom as permanent because it kept getting renewed. For them, August 20 will feel like a cut even measured against a genuinely higher floor.

What to do before August 19

Do not wait for the fifth extension to find out if it happens. Pull your actual weekly usage from Settings > Usage now, not your impression of it, and compare a boosted week against what the pre-May baseline would have covered. If the gap matters, decide before the deadline, not after a session stops mid-task.

Treat Opus-class models as the expensive option and default to cheaper models for routine work. If a workflow is business-critical, do not let it depend on one promotional number. Windsurf is worth pricing out as a second seat. Tabnine is worth a look if a self-hosted model fits the workload better than a shared quota ever will. Autonomous agents like Devin are a different bet entirely, priced by completed task rather than by session, which is its own kind of hedge against a weekly cap.

Anthropic has extended this boost four times since May without saying why a fifth would be different. August 19 is when teams that built their week around it find out whether that pattern was generosity or just an accident of scheduling.

Frequently asked questions

Did Anthropic fix the Claude Max usage limits that were running out midweek?

Anthropic doubled Claude Code's five-hour usage limits and removed weekday peak-hour throttling on May 6, 2026, for Pro, Max, Team, and seat-based Enterprise plans. The change was funded by new compute capacity from a deal for SpaceX's Colossus 1 data center. It addressed the specific complaint that Max users were exhausting their weekly budget by Wednesday.

What is the 50% Claude Code weekly limit boost, and when does it end?

It is a promotional increase to Claude Code's weekly usage limit, running on top of the May 6 baseline increase. Anthropic has extended it four times since it began, most recently pushing the end date to August 19, 2026. It applies to Pro, Max, Team, and legacy seat-based Enterprise accounts.

What happens to Claude Code limits after August 19, 2026?

Unless Anthropic extends the promotion a fifth time, weekly limits return to their standard, non-boosted level on August 20, 2026. That baseline still includes the permanent May 6 doubling of five-hour limits, so it remains higher than the limits in place before May, even without the extra 50%.

Can I pay for extra Claude usage instead of waiting for a reset?

Yes. Pro, Max 5x, and Max 20x plans can switch to pay-as-you-go usage credits billed at standard API rates once a plan limit is hit, up to a $2,000 daily cap, according to Anthropic's help center. Pro and Team Standard users on the Fable 5 model separately get a one-time $100 transition credit before metered pricing applies.

What should heavy Claude Code users do before the August 19 deadline?

Check actual weekly usage in Settings > Usage and compare it against the pre-boost baseline rather than guessing. Default to cheaper models for routine tasks and reserve Opus-class models for work that needs them. Teams with business-critical workflows should price a second option, such as Windsurf or a self-hosted model through Tabnine, rather than depending on a promotional limit.

Covered in this guide

• Anthropic: Anthropic, founded 2021 by 7 ex-OpenAI researchers, builds Claude and was valued near $965B after its May 2026 Series H round.
• Claude: Claude API with Sonnet 5 (agentic, near-Opus performance at half the cost) and Claude Science (auditable research workbench). Free to Pro.
• Claude Code: Claude Code by Anthropic scores 88.6% on SWE-bench Verified, starts at $20/month with no free tier. Reads 1M-token codebases, edits files, runs tests, and opens PRs across 8 platforms.
• Cursor: Cursor is an AI code editor built on VS Code, used by 64% of Fortune 500 companies, with Agent Mode, Tab completion, and Cloud Agents at $20/month.
• Devin: Devin is Cognition's autonomous AI software engineer that plans, codes, tests, and ships PRs from $20/month plus per-task Agent Compute Unit billing.
• Tabnine: Enterprise AI coding assistant with privacy-first architecture, agentic workflows, and flexible deployment
• Windsurf: Windsurf is an agentic AI IDE by Cognition AI featuring Cascade agents, Codemaps, and integrated Devin cloud workflows — used by developers in 70+ languages, starting free with Pro at $20/month.

Sources

• What is the Max plan?
• Manage usage credits for paid Claude plans
• Usage limit best practices
• Introducing Claude Sonnet 5
• Anthropic taps SpaceX's Colossus-1 data center for 220,000 GPUs to power Claude
• Claude Code users keep 50% higher limits until July 19
• Claude Code Limits +50%, Fable 5 Lands in Max and Team
• Anthropic finally explains why Claude usage limits feel tighter for some users