Skip to content
Call wiki

What do Claude's plan tiers buy, and when does overage beat upgrading?

Last updated 2026-08-17 · Confidence: documented — Claude’s help center, pricing page, and Claude Code cost docs; the exact subscription-vs-overage multipliers are inferred estimates, not published.

Overage almost never beats upgrading: extra usage bills at full API list rates, while a subscription’s allowance delivers the same tokens at a steep, unpublished discount — so buy the next tier (or a second subscription) before turning credits on.

The tiers

Section titled “The tiers”

Pro is $20/month ($17 annual); Max 5x ($100) and Max 20x ($200) give 5x and 20x Pro’s per-session usage. Team seats are $25/month standard, $125 premium (5x standard); annual billing drops both. All meter the same way: a rolling 5-hour session window plus a weekly cap shared across models. Verified 501(c)(3)s, K–12 schools, and qualifying healthcare orgs get Team at $8 and Enterprise at $10 per user, minimum 2 seats.

Overage: off by default, API-priced

Section titled “Overage: off by default, API-priced”

Usage credits let Pro/Max users keep working past the limit — enabled manually under Settings > Usage, prepaid, capped monthly, and billed at standard API rates. On Team and Enterprise only org admins can enable them; a member who hits the limit sees no purchase button, at most a request-to-admin link. Credits are worse than API-priced in long sessions: prompt-cache lifetime drops from 1 hour to 5 minutes once you’re drawing on them.

How much cheaper is the subscription?

Section titled “How much cheaper is the subscription?”

No official number exists. Anthropic’s own anchor: average enterprise Claude Code use runs $150–250 per developer-month at API rates — roughly a $100–200 Max plan, before you even hit limits. The multiple grows with how hard you push the allowance, which is why estimates voiced on calls ranged from ~4x through 3–5x to 5–10x. Routing bounded jobs to cheaper models shifts the math further — see API task routing.

Checking your burn

Section titled “Checking your burn”

The model can’t see your account’s limits, so asking Claude mid-chat doesn’t work. What does: claude.ai’s Settings > Usage bars, and Claude Code’s /usage command — plan-usage bars plus a breakdown attributing recent usage to skills, subagents, and MCP servers.

Q&A from calls

Section titled “Q&A from calls”

When you hit your limit, is there a “buy extra usage” button? On personal Pro/Max — only after you’ve enabled usage credits in Settings > Usage. On a Team seat there is none: the admin has to turn credits on, which is why hitting a Team limit offers only “go back or try again.”

Does “ask Claude how much of my rate limits am I eating” work? Not as a chat question — run /usage in Claude Code or open Settings > Usage instead.

Sources

Section titled “Sources”