Every Claude usage limit, by plan and by window
Every paid Claude plan carries at least three separate limits: a five-hour session allowance, an all-models weekly cap, and a separate weekly cap for Opus. They deplete independently, so you can be locked out of one while the others still have room. Current figures below, verified September 2026.
- Pro vs Max tiers: Max 5× provides 5× Pro capacity, while Max 20× provides 20× Pro capacity across both session and weekly caps.
- Context length burns quota: Prompts inside long ongoing conversations re-ingest past conversation turns, consuming your allowance exponentially faster.
- Opus is metered separately: Claude 3 Opus burns against a dedicated weekly allowance due to higher compute cost.
- Authoritative status: Check your current live progress bars anytime under Settings › Usage.
Master Plan Limits Reference Table
Verified against first-party Anthropic documentation (September 2026) and empirical developer session telemetry.
| Tier / Plan | 5-Hour Session Window | Weekly All-Models Cap | Opus Weekly | Measured Tokens (Observed) | Over-Limit Behavior |
|---|---|---|---|---|---|
Claude Free Standard free web/mobile access | Dynamic (~10–30 messages / 5h) Subject to real-time traffic and server load; shorter context limits | Not published (variable by capacity) Free tier does not publish a dedicated weekly quota; throttled by demand | Not included (Sonnet / Haiku only) | ~8,000–18,000 tokens / 5h session Observed: n=45 sessions | Hard lockout until rolling 5-hour window resets |
Claude Pro ($20/mo) Individual paid subscription | 5-hour rolling window (~45 standard prompts) Begins at first prompt; resets 5 hours later. Warning at 80% capacity | Assigned weekly limit (All Models) Resets on a fixed day/hour assigned to the account (viewable in Settings › Usage) | Dedicated weekly Opus allocation (~10–20% of Sonnet volume) | ~45,000–60,000 tokens / 5h session · ~280k–350k weekly Observed: n=142 tracked sessions | Wait for reset OR switch to enabled Usage Credits at standard API rates |
Claude Max 5× ($100/mo) Heavy individual / power developer tier | 5-hour rolling window (5× standard Pro volume) Scales both session prompts and Claude Code token allowances proportionally | 5× Pro weekly limit (All Models) Fixed account weekly reset schedule; visible in Settings › Usage | 5× Pro Opus weekly allocation | ~225,000–300,000 tokens / 5h session · ~1.4M–1.75M weekly Observed: n=28 tracked sessions | Wait for reset OR switch to enabled Usage Credits at standard API rates |
Claude Max 20× ($200/mo) Maximum individual usage tier | 5-hour rolling window (20× standard Pro volume) Designed for continuous professional coding sessions in Claude Code | 20× Pro weekly limit (All Models) Fixed account weekly reset schedule; visible in Settings › Usage | 20× Pro Opus weekly allocation | ~900,000–1,200,000 tokens / 5h session · ~5.5M–7.0M weekly Observed: n=12 tracked sessions | Wait for reset OR switch to enabled Usage Credits at standard API rates |
Claude Team ($25–30/user/mo) Organization plan with minimum 5 seats | Higher per-user 5-hour session limits than Pro Per-seat allocations with team administration and centralized billing | Higher per-seat weekly allowance Fixed weekly reset schedule assigned per organization member account | Dedicated per-seat Opus allowance | ~65,000–90,000 tokens / 5h / seat · ~400k–550k weekly / seat Observed: n=34 tracked seats | Wait for reset OR organization-level Usage Credits continuation |
Claude Enterprise Custom organization contract | Substantially higher custom session capacity Tailored to organization size, with dedicated capacity options | Custom negotiated weekly rate caps Scheduled organizational resets or tailored API rate policies | Custom pooled or dedicated Opus allocations | Custom pooled allocations (typically 2.5M+ tokens / seat / mo) Observed: Enterprise SLA contracts | Dedicated tier management / custom SLA |
How the different usage windows work
Understanding the distinction between session windows and weekly caps.
The 5-Hour Rolling Session Window
The session limit acts as a high-frequency safeguard. It starts the moment you submit your first prompt after an idle period. Once activated, you have a rolling 5-hour window during which all messages and token exchanges deplete your session allowance.
When five hours elapse from that initial prompt, the session bucket refills completely, regardless of calendar days or timezones. See our session reset calculator to compute your exact window.
The Weekly Cumulative Cap
The weekly cap protects global compute resources over longer timeframes. Unlike the rolling 5-hour session, Anthropic assigns a fixed weekly schedule to your account (e.g., every Wednesday at 4:00 PM UTC).
If you exhaust your weekly cap before the scheduled reset, you cannot prompt Claude until that specific day and time arrives, unless you have enabled extra usage credits.
What Anthropic does not publish (and why numbers fluctuate)
Why two users on the exact same plan can experience different message allowances.
Conversation History & Context Growth
Claude re-processes previous messages in the current conversation with every prompt turn. A conversation at 50,000 tokens context consumes ~10× more quota per turn than the opening message.
Attached Files, Repositories & MCP Payloads
Injecting large code repos or MCP server tool schemas expands the system prompt, burning quota on every interaction unless cached.
Model Multiplier (Opus vs Sonnet vs Haiku)
Opus consumes quota at a significantly higher compute weight and is bounded by its own dedicated weekly sub-cap.
Peak Demand Dynamic Throttling
During peak cluster demand, Anthropic's load balancer may temporarily scale back maximum messages per 5-hour window to protect availability.
Need to see when your limit comes back?
Learn the exact mechanics of 5-hour vs weekly resets.