2026 RATE-LIMIT & QUOTA REFERENCE

2026 AI Coding Subscription Quotas & Reset Master Cheat Sheet

One unified reference for 5-hour rolling windows, 7-day weekly pools, Midnight PT daily resets, and API rate tiers across Claude, OpenAI Codex, Google Gemini / Antigravity, and Cursor.

1. Anthropic Claude & Claude Code Subscription & API Quotas

View Claude Live Reset Countdown →

Claude subscriptions enforce a dual-layer limit: a 5-hour rolling window (triggered by your first prompt) plus a 7-day rolling weekly cap. Opus 4.6 consumes roughly 3–5x the window weight of Sonnet 4.6.

Plan / TierMonthly Price5-Hour Rolling Window7-Day Weekly CapReset Mechanism & Priority
Claude Free$0~10–15 short messages (load-dependent)Subject to dynamic daily/weekly throttling5h rolling after 1st prompt; Web/Mobile only
Claude Pro$20 / mo1x baseline (~45 web msgs or 10–40 Claude Code tool turns)~40–80 hrs/week equivalent Sonnet 4.6 usage+5h00m from 1st prompt; includes Claude Code
Claude Max 5x$100 / mo5x Pro budget (~225 msg-equivalents / 50–200 Agent turns)~140–280 hrs/wk Sonnet + 15–35 hrs/wk Opus+5h00m rolling + 7-day pool; supports Extra Usage overflow
Claude Max 20x$200 / mo20x Pro budget (~900 msg-equivalents / full-day multi-agent)~240–480 hrs/wk Sonnet + 24–40 hrs/wk OpusHighest queue priority; +5h00m rolling + 7-day pool
Anthropic API (Tier 1–4)Pay-as-you-goNo 5h lockout (governed by RPM / ITPM / OTPM token buckets)Controlled by monthly workspace Spend LimitPer-minute token bucket refill; cached tokens bypass standard ITPM cap

2. OpenAI ChatGPT & Codex CLI Quotas & Reset Schedule

View Codex Live Reset & Promo Status →

OpenAI Codex (CLI and cloud container tasks) shares your ChatGPT subscription’s 5-hour burst window + 7-day rolling weekly pool, alongside periodic official rate-limit resets and 2x promos.

Plan / TierMonthly Price5-Hour Window Quota7-Day Weekly Pool & Cloud TasksReset & Overflow Rules
ChatGPT Plus (Codex)$20 / mo~30–150 local CLI turns (varies by reasoning effort & context)Strict 7-day rolling cap (~2–3 heavy coding days per week)5h rolling + 7d rolling release; expandable via Codex Credits
ChatGPT Pro (Codex)$200 / mo~6–10x Plus 5-hour burst capacity (~300–1,500 CLI turns)High-capacity weekly pool supporting full-week multi-repo work5h + 7d auto-rolling; official incident grants often zero out usage
OpenAI API (Tier 1–5)Pay-as-you-goGoverned by per-minute RPM and TPM bucketsNo 7-day lockout (bounded only by credit balance & budget)Automatic prefix caching (≥1024 tokens) gives 50%–75% input discount

3. Google Gemini CLI, AI Studio & Antigravity Quotas

View Google Midnight PT Countdown →

Google combines per-model independent 5-hour rolling buckets (in Antigravity) with a hard daily RPD reset at Midnight Pacific Time (15:00 UTC+8 during PDT / 16:00 UTC+8 during PST).

Product & TierPriceRate & Window Caps (RPM / RPD / 5h)Model Pool IsolationExact Reset Schedule
Gemini CLI / AI Studio Free$0~5–15 RPM; ~25–1,000 RPD (strict on Pro, generous on Flash)Can fallback to Flash models when Pro RPD is exhaustedDaily at 00:00 PT (15:00 UTC+8 PDT / 16:00 UTC+8 PST)
Google AI Pro / Developer$19.99 / moDynamic 5-hour rolling window + elevated daily RPD poolSeparate buckets per model family inside Antigravity5-hour rolling recovery + Midnight PT daily master reset
Google AI Ultra$249.99 / moHighest 5-hour concurrent budget & priority inference laneHigh-ceiling independent pools across all model familiesContinuous 5-hour rolling replenishment
Gemini API Paid Tier 1+Pay-as-you-go150–2,000+ RPM; 2M+ TPM; high RPD ceilingIndependent RPM/TPM per model endpointPer-minute sliding RPM/TPM; 75% Context Caching discount

4. Cursor IDE Subscription Quotas & Fast Request Multipliers

Read Cursor 500 Fast Requests Guide →

Cursor uses a monthly billing-date reset pool (500 Fast Requests / $20 included usage credit); enabling Thinking or Max Mode multiplies consumption per turn.

Plan / TierMonthly PriceIncluded Fast QuotaBehavior After ExhaustionReset Cycle
Cursor Hobby (Free)$0~50 slow/trial requests + 2,000 completions per monthHard block until upgrade or next cycleMonthly on account creation date
Cursor Pro$20 / mo500 Fast Requests (or $20 equivalent frontier credit pool)Falls back to unlimited Slow Pool queue or Usage-Based billingMonthly on your billing renewal date
Cursor Ultra$200 / mo20x Pro frontier model pool + priority inference routingSupports seamless Usage-Based overflow billingMonthly on your billing renewal date
COST SIMULATOR

AI Coding Token & Cache Cost Calculator

Compare API pay-as-you-go vs $20/$100/$200 subscriptions using your actual daily token volume and cache hit ratio.

CONFIG TEMPLATES

High-Cache-Hit CLAUDE.md & AGENTS.md Templates

Copy battle-tested static prompt prefixes and 3-stage refactoring checkpoints to stretch your 5-hour window by 2.5x.

429 TROUBLESHOOTING

429 & Quota Exceeded Error Troubleshooting Wiki

Immediate recovery commands and fallback routing when you hit Claude 5-hour lockouts, Codex weekly caps, or Gemini 429s.

Frequently Asked Questions

Why can’t AI coding subscriptions be compared purely by monthly price?

Each provider enforces a fundamentally different rate-limiting geometry: Anthropic Claude uses a dual 5-hour rolling window + 7-day cumulative cap; OpenAI Codex uses a 5-hour burst window + 7-day weekly pool; Google Gemini & Antigravity use per-model 5-hour buckets plus a daily Midnight Pacific Time reset; and Cursor uses a monthly 500 Fast Request pool with slow-queue fallback or usage-based billing.

What is the real capacity gap between Claude Pro ($20), Max 5x ($100), and Max 20x ($200)?

Claude Pro ($20) provides roughly ~45 standard messages (or ~10–40 heavy tool-calling turns in Claude Code) per 5-hour window. Max 5x ($100) scales that 5-hour token weight budget by 5x (~225 message-equivalents), while Max 20x ($200) scales it by 20x (~900 message-equivalents) with priority queue access during peak hours.

When do Google Gemini CLI and Antigravity daily quotas reset?

Google AI Studio Free Tier, Gemini CLI daily request limits (RPD), and Antigravity daily model pools reset at Midnight Pacific Time (00:00 PT)—which corresponds to 15:00 UTC+8 during US Daylight Saving Time (PDT) and 16:00 UTC+8 during Standard Time (PST).

Where does Prompt Caching make the biggest difference in quota longevity?

Both direct API billing (Anthropic, OpenAI, DeepSeek) and token-weighted CLI windows (Claude Code, Codex CLI) benefit massively from Prompt Caching. Keeping `cache_read_input_tokens` above 85% cuts effective API spend by 70%–90% and stretches 5-hour rolling window endurance by 2.5x or more.