The short answer
The short answer
Pro is enough if Claude Code is one tool among several and you work in bursts. Max earns its price in one of two situations: your five-hour window runs out while paid product work is still queued, or you need Opus 5 as the working model rather than an occasional switch. Those are different reasons, and only the first one is a capacity argument.
The under-discussed half is the model default. Claude Code resolves its default model to Sonnet 5 on Pro and Team Standard, and to Opus 5 on Max, Team Premium, Enterprise and the Anthropic API. On plans, that is the difference you feel first — before you ever reach a limit.
What the plans cost and what they change
| Plan | Price | Session allowance | Claude Code default model |
|---|---|---|---|
| Free | No charge | Standard limits on a 5-hour rolling window | No Claude Code |
| Pro | $20/mo, or $17/mo billed annually | At least 5x Free per 5-hour session, plus a weekly limit | Sonnet 5 |
| Max 5x | $100/mo | Five times Pro’s per-session allowance | Opus 5 |
| Max 20x | $200/mo | Twenty times Pro’s per-session allowance | Opus 5 |
| Team Standard | $25/seat/mo, or $20 annually | Above Pro capacity, per seat | Sonnet 5 |
| Team Premium | $125/seat/mo, or $100 annually | 5x a standard seat | Opus 5 |
Max 20x costs ten times Pro and gives twenty times the per-session allowance, so on raw capacity per euro it is the better deal of the two Max tiers — provided you would actually consume it. Max 5x costs five times Pro for five times the allowance, which means you are paying for the model default and the headroom, not for a volume discount.
One allowance covers everything. Usage is shared across claude.ai, Claude Code and Cowork, so a long chat session and a repository task draw on the same budget. On Team and Enterprise plans each member’s Claude Code usage draws from their own seat allowance on the same rolling five-hour and weekly windows.
Two windows, one allowance
Every paid plan meters two ways. The session limit resets every five hours on a rolling basis. On top of that sits a weekly limit that applies across all models and resets at a fixed time each week assigned to your account — your reset day and time stay the same regardless of when you started using Claude or when the subscription began, and you get the full weekly allowance each cycle. You can see the next reset in Settings › Usage, or with /usage inside a session.
| Message you see | Which ceiling it is | What actually helps |
|---|---|---|
| "You’ve hit your session limit" | The five-hour window on your plan, shared across all models | Wait for the reset, turn on usage credits, or move up a plan. Switching model with /model does not restore access. |
| "You’ve hit your weekly limit" | The weekly window, shared across all models | Same three options. The message states when the window resets. |
| "You’ve hit your Opus limit" | A model-family ceiling, not the plan window | Switch to a model outside that family with /model and keep working. |
| "You’ve hit your monthly spend limit" | A usage-credits spend cap you or an admin set | Raise the cap, or wait for the plan reset the message names. |
| A context or auto-compact warning | Not a usage limit at all | A conversation-length problem. Clear between tasks, or set a smaller auto-compact window. |
That last row is the most common false alarm in this whole decision. People upgrade to Max because a long session kept stalling, when the actual problem was one conversation carrying an entire day of context. Claude Code usage limits walks through telling those apart before you change anything.
Why a session drains faster than your typing suggests
Claude Code sends the full conversation with every request, and each tool call sends another request carrying that batch of results. Prompt caching means the history is re-read at the cached rate rather than the full rate, but it is still read. A one-line question in a session that has been open all day still draws usage for the whole conversation.
- Cache misses: the first message after a break longer than the cache lifetime reprocesses your full context. The lifetime is an hour on a subscription and five minutes once you are drawing on usage credits.
- Agent teams: approximately 7x the tokens of a standard session when teammates run in plan mode, because each teammate holds its own context window.
- Scheduled tasks and idle check-ins: these fire on their interval and send your full context even while you are not typing.
- Compaction: /compact reads the conversation it summarises, so compacting a large context is itself a large request. /clear costs nothing.
On Pro, Max, Team and Enterprise plans, /usage shows a breakdown attributing recent usage to skills, subagents, plugins and individual MCP servers, and flags any behavior — long context, cache misses — that accounts for 10% or more of it. Read that breakdown before concluding you need more capacity: a plugin or an MCP server eating a third of your window is a configuration fix, not an upgrade.
A decision table for sustained agent work
| Your week looks like this | The plan that fits | Why |
|---|---|---|
| A few repository tasks a week, one product, Claude is one tool among several | Pro, $20/mo | Sonnet 5 handles most coding tasks and costs less. You will rarely see the weekly window. |
| Daily sessions, one product, occasional Opus for architecture | Pro plus usage credits | Credits bill at standard API rates only when you exceed the plan, with a monthly cap you set. |
| Several sessions a day, and you want Opus 5 as the working model | Max 5x, $100/mo | You are buying the default model and headroom, not a volume discount. |
| Parallel agents most of the day, or agent teams | Max 20x, $200/mo | Twenty times the allowance for ten times the price, and agent teams use ~7x the tokens of a normal session. |
| Two or more people on the same codebase | Team seats, Standard or Premium | Per-seat allowances, shared admin controls, spend reporting and Cowork. |
| Heavily automated, scripted or CI-driven runs | API billing via the Console | Anthropic’s own enterprise figure is ~$13 per developer per active day, $150–250 per month, under $30 per active day for 90% of users. |
Note where the last two rows meet. Anthropic’s own enterprise band of $150–250 per developer per month straddles Max 20x at $200, which is the honest break-even for one person coding daily: below that band a subscription is cheaper than API billing, above it the API is cheaper and gives you per-user reporting and workspace spend limits. The subscription audit is the same exercise applied across every agent you pay for.
Usage credits are the third option
Before upgrading, look at credits. Pro, Max 5x and Max 20x subscribers can turn on usage credits to keep working past the included limit, billed at standard API rates. You set a maximum monthly spend, optionally auto-reload below a threshold, and see the spend against that limit as a row in /usage. Manage it with /usage-credits after signing in with your subscription, or in Settings › Usage on claude.ai.
Credits are the right answer when overflow is occasional and bounded — one heavy week a month, a migration, a launch. A plan upgrade is the right answer when the overflow is your normal week, because credits at API rates will pass the price difference quickly.
Use a two-week decision log
Record interruptions for a normal working week, then a second one if the first was unusual. Count blocked sessions rather than total prompts, separate coding usage from general Claude use, and note whether another agent could have continued the task. This avoids buying capacity on the strength of one unusually heavy day.
Date and time:
Which ceiling (session / weekly / model family / spend):
Reset time the message showed:
Task that was interrupted:
Could another agent have continued it:
Minutes genuinely blocked:
/usage breakdown flagged (long context, cache misses, MCP, plugin):
Useful work completed in the meantime:- Count only time actually lost, excluding review you would have done anyway.
- Re-read the /usage attribution rows before blaming capacity — a plugin or MCP server can dominate a window.
- Separate a recurring capacity problem from one oversized task or a service incident.
- Re-evaluate when provider terms change; both the limits and the default models moved more than once in 2026.
Keep a second execution path
Even with more capacity, a portfolio should not depend on one provider being reachable. Keep verified local commands and a second capable agent for independent tasks — Codex, Grok CLI, Kimi CLI, OpenCode or Cursor CLI all meter separately from your Anthropic plan. Parallel capacity is only useful when the repository and the review process stay controlled; running two agents in parallel worktrees is the safe shape for it.
CmdBrief shows the windows that providers expose — Claude Code 5-hour and weekly, Codex weekly, Grok CLI weekly credits, OpenCode Go rolling, weekly and monthly, Cursor CLI monthly pools — side by side with the agent that needs you. It cannot raise an allowance, change a reset, or tell you your final invoice; those stay in the provider’s own interface.
Questions
Is Claude Max required for Claude Code?
No. Claude Code is included with Pro at $20 a month. Max changes the per-session allowance and the default model: Claude Code defaults to Sonnet 5 on Pro and Team Standard, and to Opus 5 on Max, Team Premium, Enterprise and the Anthropic API.
What is the difference between Max 5x and Max 20x?
Price and per-session allowance only: $100 a month for five times the Pro allowance, $200 for twenty times. Both default Claude Code to Opus 5, and both use the same five-hour and weekly window mechanics.
When does the weekly limit reset?
At a fixed time each week assigned to your account. It does not move with your subscription date or when you start using Claude, and you receive the full weekly allowance each cycle. Check Settings › Usage or run /usage in a session.
Should I buy usage credits instead of upgrading?
Credits suit occasional, bounded overflow; they bill at standard API rates with a monthly spend limit you set. If you exceed the plan in a normal week, an upgrade is usually cheaper — and note that while credits are in use the prompt cache lifetime drops from an hour to five minutes.
When is upgrading worthwhile?
When repeated session or weekly interruptions cost more in delayed product work than the plan difference, or when you want Opus 5 as the working default rather than an occasional switch. Log two weeks of interruptions before deciding.
Still unsure? Write to [email protected] or use the contact form. Contact CmdBrief →
Primary sources
Provider terms and product behavior can change. Confirm plan, pricing, model, and permission details in the current official documentation.