Blog · 2026-08-24 · Vynaris Team
Claude Code's 50% weekly-limit boost ends August 31: capacity costs rise 50%
Use over 66.7% of Claude Code's promo weekly limit? The same work will not fit after August 31. Capacity-unit cost rises 50%.
If you now consume more than 66.7% of Claude Code's promotional weekly limit, the same workload will not fit after August 31. Anthropic will cut included weekly capacity from an indexed 150 units to 100 while billing stays unchanged. Effective cost per included unit therefore rises 50%. Prices verified 2026-08-24.
TL;DR
- Anthropic's 50% weekly-limit promotion ends 2026-08-31 at 11:59 PM PT. Standard weekly limits return after that deadline.
- The reset removes one-third of promotional weekly capacity:
(150 - 100) / 150 = 33.3%. - Fixed subscription prices make each remaining capacity unit 50% dearer than during the promotion:
(P / 100) / (P / 150) - 1 = 50%. - The five-hour usage limit does not change. Only the weekly ceiling moves.
Verdict table
Current use of promotional weekly limit Demand versus standard limit Workload cut needed to fit Verdict
--------------------------------------- ---------------------------- -------------------------- --------------------------------------
50% 0.75x 0% No weekly-limit impact
66.7% 1.00x 0% Exact planning boundary
70% 1.05x 4.8% Trim or shift a small tail
80% 1.20x 16.7% Re-plan before September
100% 1.50x 33.3% One-third of current work will not fitThe table uses relative capacity, not tokens. Anthropic does not publish a fixed token budget for this offer. Calling 150 units “150 million tokens” would be fiction.

What changes on September 1
Anthropic says eligible Pro, Max, Team, and legacy seat-based Enterprise users receive 50% more weekly Claude Code usage from May 13 through August 31. The promotion applies in the CLI, IDE extensions, desktop app, and web. Free users and consumption-based Enterprise seats are excluded.
After the deadline, the plan and billing stay the same. Weekly usage returns to its standard level. The separate five-hour limit stays where it is throughout.
That distinction matters. A coding agent can hit the five-hour ceiling even when plenty of weekly capacity remains. The promotion never relaxed that shorter window. September only changes the weekly denominator.
The denominator math
Let the standard weekly allowance be C. The promotion makes it 1.5C. Let the fixed plan price be P.
promotional cost per capacity unit = P / 1.5C
standard cost per capacity unit = P / C
post-promotion increase = (P/C) / (P/1.5C) - 1 = 50%
capacity removed = (1.5C - C) / 1.5C = 33.3%Those two percentages describe opposite directions. Capacity falls 33.3% from the larger promotional base. Effective cost rises 50% from the cheaper promotional base. Neither calculation changes the sticker price.
For Anthropic's current Max tiers, the normalized result looks like this:
Plan Monthly price Promo cost per baseline-capacity unit Standard cost per baseline-capacity unit
------- ------------- ------------------------------------- ----------------------------------------
Max 5x $100 $66.67 $100.00
Max 20x $200 $133.33 $200.00One “baseline-capacity unit” means one standard weekly allowance, whatever Anthropic's internal meter grants that account. It is not a claim about messages, tokens, or completed tasks. Our Claude Max versus API break-even analysis uses the same restraint: public limits support relative comparisons, not a guaranteed session count.
The 66.7% planning threshold
Suppose /usage says you consumed share s of the promotional weekly limit. Current demand is s × 1.5C. It fits the standard limit only when:
s x 1.5C <= C
s <= 2/3
s <= 66.7%That is the actionable threshold. A team below two-thirds can keep the same weekly workload, subject to the unchanged five-hour limit. A team above it must reduce demand, move work, or accept interruptions.
At 80% promotional usage, demand equals 0.8 × 1.5C = 1.2C. Cutting from 1.2C to C means removing 0.2 / 1.2 = 16.7% of today's workload. At full promotional usage, the required cut is 33.3%.
Check /usage, then use the cost calculator to price the overflow on an API route with your real input and output counts. The calculator cannot predict Claude Code's unpublished allowance. It can price work you deliberately move outside the subscription.
What to do before the reset
First, record daily /usage readings through one representative workweek. This is capacity planning, not cost per token accounting. The product meter is the only honest denominator available for your account.
Second, separate interactive work from automatable runs. Keep human-in-the-loop debugging in Claude Code. Move repeatable tests, batch transforms, and scheduled jobs to an API only when their cost per task and failure handling are measurable.
Third, reserve capacity for work that cannot wait. A hard cost ceiling is useful, but a weekly cap reached on Thursday is still an operational failure. Our hard spend-cap guide shows the same principle for metered APIs: reserve before dispatch, then reject or defer work explicitly.
Honest tradeoff
Do not move work to the API merely because your promotional usage exceeds 66.7%. Subscription capacity is bundled, while API calls create a marginal bill and require credentials, logging, retries, and spend controls. Trimming low-value agent loops may cost less than building a second execution path.
The reverse is also true. If a weekly interruption blocks releases, fixed-price convenience has failed its job. A metered fallback can be rational even when its unit price is higher. Compare the cost of delay with the API invoice, not with zero.
FAQ
Does the Claude Code subscription price increase after August 31? No. Anthropic says there is no change to the plan or billing. The included weekly limit returns to its standard level.
Why does effective cost rise 50% when capacity falls 33.3%? The bases differ. Capacity moves from 150 to 100, a one-third fall. Cost per unit moves from P/150 to P/100, a 50% rise.
Will the five-hour usage limit also shrink? No. Anthropic says the five-hour limit is unaffected by the promotion and its expiry.
Which plans lose the promotional capacity? Pro, Max, Team, and legacy seat-based Enterprise users receive the temporary increase. Free plans and consumption-based Enterprise seats do not.
Can this be converted into an exact token allowance? No. Anthropic publishes the relative increase, not a fixed token amount. Use /usage for account-specific planning.
Sources
- Anthropic: Claude Code May-August 2026 weekly limits promotion, eligibility, dates, scope, unchanged five-hour limit, and billing behavior verified 2026-08-24.
- Anthropic: What is the Max plan?, Max 5x at $100 per month and Max 20x at $200 per month verified 2026-08-24.
- Reproducible arithmetic:
artifacts/claude-code-50-percent-weekly-limit-promo-ends-august-31-math.py.