Guide
Claude Pro's $20 Wall: Wait, Upgrade to Max, or Buy Usage Credits?
Claude Pro is a usage envelope, not unlimited access. Compare a five-hour reset, Max from $100, and separately billed usage credits without creating a surprise bill.
Claude Pro is not an unlimited seat. The monthly plan shares usage between Claude chat and Claude Code. When included usage is exhausted, a paid-plan user can wait for the normal five-hour reset, move to Max from $100 per month, or deliberately enable prepaid usage credits billed at standard API rates.
The cost-only decision is straightforward:
- Wait when limits are occasional and the pause is workable.
- Use capped credits when overflow is rare and measured monthly credit spend stays below the upgrade gap.
- Evaluate Max 5x when overflow is the baseline or Claude Fable 5 is regular work, then confirm its limited Fable allowance covers the workload.
Prices and credit rules below were checked on 28 August 2026. Confirm the current plan and API rate before purchasing.
What Pro and Max currently include
Anthropic’s pricing page lists Pro at $20 per month on monthly billing. The annual option is $200 up front, shown as $17 per month after the annual discount. Pro includes Claude Code, Cowork, Design, Science, Research, unlimited projects, and access to more models.
The same page lists Max from $100 per month, with a choice of 5x or 20x more usage than Pro. Anthropic clarifies that this is usage per five-hour session and that other weekly, monthly, model, or feature caps may still apply.
Fable access is the important exception. The comparison table lists Fable on Pro as Usage credits. Max 5x and Max 20x list Fable as 50% of weekly limits. A Pro user should therefore not assume there is a monthly Fable allowance to exhaust before credits begin.
Claude Code and chat share the same plan usage. Anthropic’s pricing FAQ says terminal work and chats draw from one pool, while the usage-credit guide says credits apply across Claude conversations and Claude Code.
How usage credits work
Anthropic’s help page gives an explicit activation flow: open Settings > Usage, locate Usage credits, click Enable, configure payment and spending preferences, set a monthly limit, and add prepaid funds. Auto-reload is optional. Credits are not a hidden part of the $20 subscription; they are charged separately.
When included session usage is reached and credits are enabled with funds available, the user can choose to continue at standard API rates. The regular session reset still occurs every five hours. Credits can be disabled later, after which only included plan usage remains available.
The page also confirms:
- A $2,000 daily redemption limit. That is Anthropic’s label; do not treat the large redemption ceiling as a sensible personal safety budget.
- The Usage dashboard separates included usage from credit consumption.
- Mobile subscribers must enable and purchase credits on the Claude web app.
- Credits usually do not expire, but in jurisdictions such as Japan they expire six months after purchase starting 10 September 2026.
- Claude gives a notification when approaching and reaching included limits before continued use switches to credits.
Set the monthly cap before adding funds. Leave auto-reload off until one billing cycle shows what the overflow actually costs.
The verified Fable 5 rates
Anthropic lists these standard Fable 5 rates per million tokens:
| Token type | Price per million tokens |
|---|---|
| Input | $10 |
| Five-minute cache write | $12.50 |
| Cache read | $1 |
| Output | $50 |
These rates are much higher than current Sonnet or Haiku rates. Use Fable arithmetic only for a Fable workload. For other models, repeat the calculation with the live rate for the model actually used.
Worked example: credits versus the $80 Max gap
Monthly Pro is $20 and Max 5x starts at $100, so the fixed upgrade gap in this cost-only comparison is $80 per month before tax. The example below is a modeling choice, not an Anthropic estimate of an hour of work.
Assumptions:
- One heavy coding turn processes 80,000 tokens of project context and produces 8,000 output tokens.
- One modeled “heavy hour” contains 12 such turns.
- The first turn writes the 80,000-token context to the five-minute prompt cache.
- The following 11 turns read that same cached context.
- The example ignores cache expiry, misses, retries, tool-side costs, long-context adjustments, tax, and regional pricing.
First-turn cost:
0.08 × $12.50 cache write + 0.008 × $50 output = $1.40
Each later turn:
0.08 × $1 cache read + 0.008 × $50 output = $0.48
One modeled heavy Fable hour:
$1.40 + 11 × $0.48 = $6.68
At that modeled rate, API spend crosses the $80 monthly price gap at roughly 12 heavy Fable hours. This is a spending threshold, not a claim that Max includes 12 of these modeled hours. Anthropic does not convert Max’s limited Fable allowance into hours, so verify capacity in the live Usage view. Because Fable on Pro is a usage-credit feature, this comparison does not subtract an invented Pro Fable allowance.
| Modeled Fable usage on Pro | Approximate API-rate credit cost | Relation to the $80 monthly price gap |
|---|---|---|
| Wait through blocked windows | $0 | No extra spend |
| 4 heavy hours per month | $26.72 | Below the price gap |
| 12 heavy hours per month | $80.16 | Approximately equal to the price gap |
| 20 heavy hours per month | $133.60 | Above the gap; check whether Max capacity covers the workload |
A 15-turn modeled session would cost about $8.12: one $1.40 cache-write turn plus 14 cache-read turns at $0.48. Four such sessions a week across an average 4.3-week month would be about $140 in credits. That is illustrative arithmetic, not a prediction of cache behavior or a claim about typical Claude Code use.
Max may also have value beyond this token comparison, including more general usage and higher output limits. Conversely, credits let a light user pay only during a rare spike. Use the $80 line as a cost threshold, then separately value predictability, model access, and the ability to wait.
Keep credits from becoming the product
Start by checking why included usage ended. Long context, Research, project files, and Code all consume tokens. Moving the same workload to credits does not make it cheaper; it only keeps it running.
Use a small prepaid balance and a monthly cap. Watch Settings > Usage during the first paid session. Disable credits after the deadline if the overflow was temporary. Anthropic recommends using included five-hour windows first, choosing efficient models when they are sufficient, starting a new conversation for a new topic, and storing recurring material in project knowledge instead of repeatedly uploading it.
Do not create additional accounts or probe reset boundaries to manufacture more included usage. Waiting, a larger plan, and explicitly budgeted credits are the supported choices.
A stale comparison to avoid
An older XDA account of cancelling Claude compared Claude’s limits with ChatGPT Plus. Treat that competitor comparison as stale: it predates the 25 August restoration of ChatGPT Plus’s five-hour Codex and Work limit. The durable Claude-side plan and credit rules are the current Anthropic pages above, not a third-party snapshot.
Decision checklist
- Wait if the wall is occasional and a five-hour reset does not endanger paid work.
- Cap and credit if overflow is rare, the live model rate is known, and a hard monthly limit stays below the value of moving to Max.
- Evaluate Max 5x if extra usage is recurring, Fable is regular work, or predictability matters more than a variable credit bill; confirm the live limits before assuming it covers the modeled hours.
- Recheck after one month using actual credit consumption rather than modeled hours.
Sources checked 28 August 2026: Claude pricing, Claude Platform model pricing, and Manage usage credits for paid Claude plans. The XDA article is retained only as a stale, third-party comparison warning.