Codex Limits Comparison: Plus / Business / Pro 100 / 200 / 500 (2026)
TL;DR: ChatGPT Plus and Codex 5-hour limits are fully back, resetting dynamically on a rolling 5-hour window with each model counted independently. On plans: Plus and Standard Business have identical Codex quotas; Pro 100 is 5x the Plus quota, Pro 200 is 10x, and Pro 500 is 25x. If quota gets tight, switch to the higher-quota model (GPT-6 Luna), use the Codex App client (double quota), or upgrade to Pro.
📚 Table of Contents (click to jump)
- 1. Are ChatGPT Plus 5-Hour Limits Back? How the Reset Mechanism Works
- 2. How Much Codex Quota Does Each Plan Get? Five Plans Across Four Models
- 3. How to Check Remaining Codex Usage: Four Official Methods
- 4. Out of Quota? Five Practical Ways to Save Tokens
- 5. Plus, Business, or Pro: Which to Choose?
- 6. Pros, Cons, and Caveats of the Quota Mechanism
- 7. Which ChatGPT Plan Fits Your Budget and Needs?
- 8. Quota and Reset FAQ
1. Are ChatGPT Plus 5-Hour Limits Back? How the Reset Mechanism Works
Short answer: yes. ChatGPT Plus and Codex both strictly run a rolling 5-hour window, the limit has not been removed, and accounts using it continuously at high intensity get a Weekly Limit layered on top depending on system load.
Here are the core calculation rules straight from the official documentation:
1. When does the 5-hour limit reset, and how is the window calculated?
Definition: the ChatGPT and Codex 5-hour limit does not reset at a fixed daily clock time (midnight, noon). It's a strict dynamic rolling window.
- Start of the clock: it begins counting from the first request you send in the current cycle.
- Reset moment: after 5 continuous hours, that batch's quota is automatically released and restored.
- Independent per model: each model's quota is calculated entirely separately. For example, your
GPT-6.1 Solmessage count can hit its 5-hour cap whileGPT-6 Lunastill runs at full capacity.
2. How much is the ChatGPT Plus 5-hour limit in dollars? Are there extra charges?
Conclusion: the ChatGPT Plus 5-hour limit is measured in request counts, and the cost is $0 — no extra charges, ever.
- Plus counts requests, not dollars: Plus is a flat $20/month subscription. Within a 5-hour window, OpenAI grants a fixed message/request range depending on the model you use (from 5 to 3,000, as the table below shows). Maxing it out within those 5 hours costs you nothing extra.
- When does it bill in dollars? Only when you use an OpenAI API Key, or have pay-as-you-go billing bound in the Codex settings, is it charged per token/dollar — and that calling method isn't constrained by the 5-hour message cap.
3. How many 5-hour limits is a GPT Plus Weekly Limit?
Answer: based on community testing and widespread developer reports, the Weekly Limit is roughly equivalent to running 10–15 full 5-hour cycles at maximum intensity within 7 days.
- How it shows up: if you trigger the 5-hour cap at high intensity for several days running, you may see the limit persist even after the 5 hours have passed.
- What to do: ordinary day-to-day development very rarely triggers the weekly limit. If you're genuinely running heavy automation or a team, upgrade to Pro 100 / Pro 200 / Pro 500, or rotate between spare accounts.
4. Do Codex and ChatGPT Work have separate quotas, or share one?
Conclusion: Codex Local Messages and Cloud Chats share the same 5-hour window, but that window is entirely independent of regular ChatGPT web chat.
| Dimension | Meaning | Current Status |
|---|---|---|
| Local Messages / 5h | Code generation requests in your local terminal or IDE, refreshed every 5 hours | Live — the core of daily use |
| Cloud Chats / 5h | Cloud conversations, fixed to GPT-6.1 Sol | Live — shares the same 5-hour window as Local Messages, usually consumes quota faster |
| Code Reviews / 5h | Automated code review | Not yet available |
💡 Key point: Local Messages and Cloud Chats share one 5-hour window, and Cloud Chats typically burns quota faster than Local Messages. OpenAI may layer on a weekly limit. API Key calls are billed by usage and don't draw on the 5-hour quota.
2. How Much Codex Quota Does Each Plan Get? Five Plans Across Four Models
Conclusion: Plus and Standard Business have identical Codex quotas (baseline 1x); Pro 100 is 5x the Plus quota; Pro 200 is 10x; Pro 500 is the 25x top tier.
Data source: the official OpenAI usage dashboard (2026, latest GPT-6 architecture series).
| Model | Plus (baseline) | Standard Business | Pro 100 (5x) | Pro 200 (10x) | Pro 500 (25x) |
|---|---|---|---|---|---|
| GPT-6 Astra | 5–45 | 5–45 | 25–225 | 50–450 | 125–1,125 |
| GPT-6.1 Sol | 15–160 | 15–160 | 75–800 | 150–1,600 | 375–4,000 |
| GPT-6 Sol | 15–150 | 15–150 | 75–750 | 150–1,500 | 375–3,750 |
| GPT-6 Luna | 350–3,000 | 350–3,000 | 1,750–15,000 | 3,500–30,000 | 8,750–75,000 |
Below are actual screenshots from the OpenAI official usage dashboard:
Reading the official data
- Plus and Standard Business have identical Codex quotas. Both show exactly the same 5-hour request counts on GPT-6 Astra, GPT-6.1 Sol, GPT-6 Sol, and GPT-6 Luna. Business's advantage isn't extra Codex quota — it's 15 bonus Pro model sessions plus consolidated invoicing and centralized team permissions.
- Pro 100 is exactly 5x Plus. Taking the everyday high-frequency model GPT-6.1 Sol, Plus gets 15–160 while Pro 100 rises to 75–800; the lightweight GPT-6 Luna reaches 1,750–15,000, enough to cover round-the-clock high-frequency coding.
- Pro 200 is exactly 10x Plus. GPT-6.1 Sol provides 150–1,600 per 5h, and GPT-6 Luna reaches 3,500–30,000 per 5h, supporting parallel work across multiple devices and deep automated builds.
- Pro 500 is the absolute ceiling at 25x. GPT-6 Luna peaks at 75,000 per cycle and GPT-6.1 Sol at 4,000. Built for heavy AI architects and fully automated agent pipelines — quota anxiety, eliminated.
- Pairing the generous model with the flagship model. Within one plan, GPT-6 Luna (350–3,000) offers roughly 20x the available count of GPT-6.1 Sol (15–160). Use Luna first for code completion, formatting, unit test generation, and comment translation; switch to Sol or Astra for hard debugging and core architecture work.
3. How to Check Remaining Codex Usage: Four Official Methods
Answer: the fastest way is to check directly in the Codex App client (which carries double quota), or visit the official web dashboard at chatgpt.com/codex/settings/usage.
Here are all four methods in full:
Method 1: Codex App Client (Recommended)
To promote the Codex app, OpenAI offers doubled quota on it. With the same plan and model, the App allows twice as many requests as the CLI and the web. Click the local button in the interface to view your remaining percentage in real time.
Method 2: Web Dashboard
Visit chatgpt.com/codex/settings/usage in your browser. Once signed in, you see your current 5-hour remaining quota and a bar chart of the distribution across models.
Method 3: CLI Terminal
Type /status in the terminal and the system immediately returns each model's remaining call quota and the refresh countdown. Ideal for keyboard-only developers.
Method 4: IDE Extension Panel
In editors like VS Code or JetBrains, click the Codex status icon in the bottom-left of the window and the panel expands to show your remaining quota and connection status.
⚠️ A warning about Fast mode: enabling Fast mode significantly improves response speed but doubles quota consumption. For non-urgent, non-parallel tasks, leaving it off is strongly recommended.
4. Out of Quota? Five Practical Ways to Save Tokens
Answer: when your main model reports You've reached your usage limit, the fastest fixes are switching to the high-quota GPT-6 Luna model or continuing in the Codex App where quota is doubled.
These five come from real day-to-day use:
- Move simple tasks to the generous model. GPT-6 Luna's quota is roughly 20x GPT-6.1 Sol's. Formatting code, generating comments, and renaming variables should run on Luna, keeping the flagship model's quota for complex refactors.
- Use the Codex App client (double quota). OpenAI grants the Codex App client double the available quota. When the web or CLI reports a limit, switching to the desktop app often lets you keep working.
- Open a new session window to clear context. The longer a session runs, the more historical tokens ride along with every request, draining quota faster. Starting a new session after finishing a module noticeably reduces token spend.
- Put AGENTS.md in a subdirectory, not the root. In a large project, nesting AGENTS.md in a specific subdirectory controls the context volume injected each time and avoids global redundancy.
- Disable MCP servers you aren't using. Every active MCP tool adds another layer of context wrapping to each message. Turn off unused MCP services promptly.
When should you insist on the strongest model? For core business architecture or complex bug debugging, use the top model (GPT-6.1 Sol or Astra) without hesitation. The cost of grinding on a bug for hours far exceeds the quota you'd save.
5. Plus, Business, or Pro: Which to Choose?
Recommendation: for everyday personal coding, Plus is enough; for team collaboration with occasional Pro model access, choose Business; for high-intensity coding, Pro 100; for deep multi-device coordination, Pro 200; for large-scale automation pipelines, Pro 500.
| Plan | Official Monthly Fee | Core Quota Multiplier | Best For |
|---|---|---|---|
| Plus | $20 | Baseline (1x) | Individual developers; daily Codex and chat are sufficient |
| Business | $25 (price adjusted) | Baseline (1x) + 15 Pro sessions | Organizations needing team permission management and occasional Pro model access |
| Pro 100 | $100 | 5x the Plus quota | Independent full-stack developers using Codex heavily |
| Pro 200 | $200 | 10x the Plus quota | Professional engineers running frequent concurrent sessions |
| Pro 500 | $500 | 25x the Plus quota | Top-tier AI power users running round-the-clock agent systems |
💡 Through upgpts.com you can open or upgrade Plus, Business, or any Pro tier directly with WeChat / Alipay, with no overseas credit card required — legitimate physical cards, activated instantly.
6. Pros, Cons, and Caveats of the Quota Mechanism
Advantages
- Transparent official data: the boundaries between 4 models × 5 plans are clearly drawn, so there's no need to guess.
- Precise tier ladder: from Plus to Pro 500 spans 25x, cleanly covering light, everyday, high-frequency, and fully automated needs.
- Multi-model, multi-client self-rescue: hitting one model's cap doesn't affect the others, and the App grants an extra 2x quota on top.
Caveats and weaknesses
- Rolling ranges, not fixed absolutes: the system floats the range with current compute load (15–160, for example) rather than using rigid numbers.
- Cloud chats burn faster: Cloud Chats and Local Messages share the 5-hour window, so heavy cloud conversation drains quota sooner.
- Fast mode doubles consumption: enabling Fast mode doubles the rate quota is deducted, so remember to turn it off.
7. Which ChatGPT Plan Fits Your Budget and Needs?
- Individual developers / light coding assistance → ChatGPT Plus ($20/month): the best value, and pairing it with Luna covers everyday needs completely.
- Small-to-mid R&D teams / collaboration needs → ChatGPT Business ($25/month): consolidated invoicing and member permission management, plus 15 bonus top-tier Pro sessions.
- Full-time independent developers / high-frequency heavy coding → ChatGPT Pro 100 ($100/month): the best balance point for long immersive coding days, with 5x the full quota.
- Tech leads / multi-device heavy power users → ChatGPT Pro 200 ($200/month): 10x quota, unfazed by concurrent multi-terminal work.
- AI labs / fully automated agent matrices → ChatGPT Pro 500 ($500/month): the 25x ceiling, saying goodbye to usage limits entirely.
8. Quota and Reset FAQ
Are ChatGPT Plus 5-hour limits back?
When do ChatGPT / Codex limits reset, and how is the 5-hour window calculated?
How much is the ChatGPT Plus 5-hour limit in dollars?
How many 5-hour limits is a GPT Plus Weekly Limit?
Do Codex and ChatGPT Work / regular chat share the same quota?
Are the Codex limits the same on Plus and Business?
How many times the Plus quota are Pro 100, Pro 200, and Pro 500?
How do I check my remaining Codex usage?
Does Fast mode consume more usage?
Planning to upgrade to ChatGPT Plus, or open Pro 100 / Pro 200 / Pro 500? See the latest top-up page for reliable official channels.