ChatGPT Codex Limits Are Back: What Changed?

OpenAI has restored the five-hour usage limit for Codex and ChatGPT Work for Plus subscribers, effective August 25, 2026. After users reported and experienced the temporary removal of this constraint for several weeks, the short-term usage window has returned, though with an important distinction: Pro $100 and Pro $200 subscriptions will not have the 5-hour limit enabled for the upcoming months.

This change matters for developers relying on intensive Codex sessions. Understanding how the 5-hour limit interacts with weekly usage caps, why tasks consume varying amounts of allowance, and what options exist when you hit a ceiling can mean the difference between a productive workflow and repeated interruptions.

Quick Answer: Are ChatGPT Codex Limits Back?

Yes. The five-hour usage limit is returning to Codex and ChatGPT Work for Plus subscribers starting August 25. The limit is a rolling window that gates how much work you can do before waiting for a reset. Usage limits for local messages and cloud tasks share a five-hour window, with additional weekly limits potentially applying. Pro higher-tier users remain exempt for now.

Key takeaway: Plus users should expect the 5-hour window to affect intensive coding sessions, but Pro $100 and Pro $200 users have additional time for concentrated work.

What Changed With ChatGPT Codex Limits?

The timeline over the past months reveals shifting constraints on developer work.

The 5-hour window was temporarily lifted

On July 12, 2026, OpenAI temporarily removed the 5-hour usage limit restriction for all Plus, Business, and Pro plans. This followed what OpenAI described as intense demand. The change followed a sharp demand spike around GPT-5.6 Sol, Codex, and ChatGPT Work. During this period, weekly usage limits continued to apply while the short-term five-hour window was absent.

The limit returns, with a plan-specific twist

OpenAI announced that the five-hour limit is returning to Plus accounts while keeping it disabled for Pro $100 and Pro $200 subscriptions for the upcoming months. The company stated that the 5-hour limit allows it to smooth the load on its compute while keeping the plan generous in terms of weekly usage, and that Plus users (described as relatively casual and new) can accidentally consume their entire weekly allowance and become confused by the restriction.

Why this matters for Plus users

A five-hour window creates practical friction for developers who prefer long, uninterrupted coding sessions. When you hit the limit, you cannot start new Codex turns until the window resets, even if your weekly allowance has capacity remaining. Developers debugging issues, refactoring large codebases, or working through iterative AI-assisted development can exhaust a 5-hour window quickly with complex tasks.

What Is the ChatGPT Codex 5-Hour Limit?

The 5-hour limit is a rolling usage window that meters how much Codex activity you can perform before you must wait for the window to reset. Usage limits for local messages and cloud tasks share this five-hour window, meaning CLI-based work and cloud-based tasks draw from the same pool.

The window is not measured in message counts. Since April 2, 2026, Codex credit pricing is tied to input tokens, cached input tokens, and output tokens rather than a simple fixed cost per message. A small CSS fix consumes far less allowance than a complex multi-file refactor with debugging and testing cycles.

The exact allowance within the window is not published as a fixed number. OpenAI does not publish fixed message caps because metering is by tokens since April 2026, so how far a window stretches depends on how heavy your requests are.

5-Hour Codex Limit vs Weekly Limit

Understanding how these two constraints interact is essential for managing your workflow.

You can hit the 5-hour window while still having weekly capacity left. Conversely, you can stay within the 5-hour window but exhaust your weekly allowance. Each constraint operates independently.

Why Does Codex Usage Run Out So Quickly?

Task complexity and scope are the primary drivers of consumption. OpenAI states usage varies with the model, context, reasoning, tool use, retrieval, caching, task duration, and whether work runs locally or in the cloud.

Task complexity: A simple task requesting a single function modification consumes minimal allowance. A request asking Codex to review a large repository, identify issues, modify multiple files, run tests, and debug failures can consume substantially more in a single session.

Large repositories and context: The size of the codebase you reference increases token usage. Large repos, long chats, and big AGENTS.md files cost more.

Long-running agentic tasks: A heavyweight refactor that drags a large repository into context spends far more than a one-line tweak, even though both are one Codex task.

Model selection: Different GPT-5.6 models consume different token counts. The efficient Luna model uses fewer tokens than the flagship Sol model.

Reasoning and tool usage: Tasks requiring extended reasoning or calling external tools consume more tokens for the model’s processing and output.

Repeated or parallel tasks: Running multiple Codex sessions in parallel or repeating similar tasks in succession can quickly exhaust an allowance.

Example: A developer asks Codex to inspect a Next.js application, identify authentication issues, modify the auth module, update several related files, run unit tests, debug failures when tests fail, and refactor based on results. This single workflow involves large context loading, reasoning, multi-file changes, tool calls, and repeated iteration. It will consume substantially more allowance than a simple request for a single bugfix.

Read More: Claude Code vs OpenAI Codex: Features, Benchmarks & Pricing

What Happens When You Hit the Codex Usage Limit?

When your usage exceeds the 5-hour window, Codex will not start new turns. If a limit lands mid-turn, Codex is allowed to finish the turn it is on, within fair-use bounds.

At this point, you have several options:

Wait for the reset. The 5-hour window rolls continuously based on when you began your usage activity. After five hours elapse, new capacity becomes available.

Check remaining usage. In the CLI, run /status in a Codex session to see how much of the 5-hour and weekly windows you have used. In ChatGPT, open the usage panel in Codex settings.

Purchase credits. ChatGPT Plus and Pro users who reach their usage limit can purchase additional credits to continue working without needing to upgrade their existing plan.

Use an available reset. Eligible users can send Codex invitations from the profile menu in the lower-left corner of the app and receive banked rate-limit resets when recipients send their first Codex message.

Upgrade the plan. Pro plans offer higher usage allowances than Plus.

How to Check Your Codex Usage and Reset Time

Direct visibility into your remaining allowance is essential for managing intensive work.

In the Codex CLI: Run the /status command within a Codex session. This displays remaining 5-hour window usage and weekly allowance consumed.

In ChatGPT: Open Codex settings and navigate to the Usage panel. This displays remaining credit balance, current limits, and per-model consumption.

What to look for: The interface shows how much of your allowance remains, the reset time for the 5-hour window, your weekly usage status, and any available banked resets or credits.

Check the usage panel regularly during intensive work sessions, especially before starting large agentic tasks or complex refactoring.

Can You Get More Codex Usage Without Upgrading?

Several options exist depending on your plan and account status.

Wait for the limit to reset

The 5-hour window rolls based on usage activity. After five hours pass, new capacity becomes available. For weekly limits, reset typically occurs on a calendar boundary (commonly weekly).

Use an available reset

Referral rewards include banked rate-limit resets usable for 30 days after they’re granted. Redeeming a reset restores both the 5-hour and weekly allowances. Check your profile for active referral promotions.

Purchase eligible Codex credits

ChatGPT Plus and Pro users who reach their usage limit can purchase additional credits to continue working. Credits are priced according to OpenAI’s token rate card. Exact pricing depends on model selection.

Upgrade the plan

Pro users receive higher usage allowances than Plus users. Pro gives either 5x or 20x higher Codex usage than Plus, depending on the tier.

Availability of credits and resets depends on your plan and account status. Enterprise and Business workspaces use separate credit systems.

How to Make Your Codex Usage Last Longer

Practical strategies help extend your allowance across productive work.

1. Give Codex focused tasks. Break work into logical chunks rather than asking Codex to handle an entire feature in one turn. A focused request costs less than an open-ended one.

2. Keep repository context relevant. Only include files Codex needs to see. Larger context windows increase token consumption.

3. Break large projects into stages. Rather than asking for complete implementation, request implementation, then testing, then optimization as separate turns.

4. Avoid unnecessary repeated prompts. Reword a request once and wait for results before retrying. Repeated variations of the same request burn allowance.

5. Let Codex inspect before expanding scope. Have Codex examine relevant files first, then request changes. This avoids reloading context unnecessarily.

6. Use lighter models when appropriate. Luna is the lighter model at $0.20/$1.20 per million tokens (down 80% after July 2026 pricing cuts), while Sol runs at $5/$30. Simple tasks benefit from Luna.

7. Check usage before large tasks. Review your remaining allowance before starting complex work. This prevents mid-task interruption.

8. Enable caching where supported. Cached input costs roughly a tenth of fresh input, which is the single biggest lever you have. Reuse context across sessions.

Does the Codex Limit Affect Normal ChatGPT Usage?

No. Codex usage and regular ChatGPT conversation are separate allowances. Regular ChatGPT chat has its own usage model distinct from Codex. Exhausting your Codex window does not restrict ChatGPT’s general conversation features.

However, ChatGPT Work and Codex share the same agentic usage and credit pool. If you use ChatGPT Work for research or analysis, it draws from the same allowance as Codex.

Is Codex Usage the Same as OpenAI API Usage?

No. They are separate billing systems.

Through ChatGPT plans: Codex access is bundled into your subscription (Plus $20/month, Pro from $100/month). You have an included allowance, then can purchase credits for overages.

Through the OpenAI API: Developers using an API key authenticate differently and receive billing through the OpenAI Platform at standard API token rates. This is completely separate from ChatGPT plan usage.

Switching to API-based Codex does not provide unlimited usage. It converts the consumption model from included allowance to per-token billing, which may be cheaper or more expensive depending on your usage pattern.

Are Pro Users Affected by the Same Codex Limit?

The five-hour limit is being restored for Plus accounts, but Pro $100 and Pro $200 subscriptions will not have the 5-hour limit enabled for the upcoming months.

This distinction is recent and plan-specific. Pro users have more flexibility for long sessions without the rolling 5-hour constraint.

What Codex Users Are Saying About the Limit Returning

Community reactions highlight practical workflow concerns:

Developers have expressed frustration with short-term interruptions during intensive sessions. Many prefer a single flexible weekly allowance over a rolling 5-hour window. Some are concerned about long uninterrupted coding sprints becoming difficult on Plus plans.

Users debate whether the 5-hour window or larger weekly allowance better serves different working styles. Questions about flexibility and whether users can customize limit behavior appear frequently. Plus users specifically ask whether the recent changes reflect planned permanent behavior or temporary adjustments.

These are community observations and represent user sentiment, not OpenAI policy.

Is the ChatGPT Codex 5-Hour Limit Permanent?

OpenAI announced the restoration is effective August 25, 2026, with the explanation that the limit smooths compute load while keeping weekly allowances generous. The company has not published an explicit end date or stated that this configuration is permanent.

OpenAI has adjusted Codex limits multiple times in 2026. The limit’s current status reflects current policy, but policies can change. Treating official documentation as authoritative rather than community estimates provides the clearest picture.

Codex Limits: What We Know vs What We Don’t

This separation helps distinguish reliable information from speculation or community reports.

Final Verdict

The ChatGPT Codex 5-hour limit has officially returned for Plus users as of August 25, 2026. OpenAI’s stated reason is that the limit smooths compute load while allowing weekly allowances to remain generous, and Plus users can accidentally consume their entire weekly allowance without understanding the restriction.

Usage limits for local messages and cloud tasks share the five-hour window, with additional weekly limits potentially applying. These are separate constraints. You need to monitor both.

Since April 2, 2026, usage is metered by tokens rather than a fixed per-message model, meaning complex tasks consume substantially more allowance than simple ones.

Your options when reaching a limit include waiting for reset, purchasing credits, using banked resets if available, or upgrading to a higher plan. The best approach is to monitor your current usage dashboard, break large tasks into focused sessions, and track both 5-hour and weekly allowances before intensive work.

Pro $100 and Pro $200 users remain exempt from the 5-hour limit for the upcoming months, offering more flexibility for concentrated coding work.

Frequently Asked Questions

Is the ChatGPT Codex 5-hour limit back?

Yes, as of August 25, 2026, the 5-hour usage window has returned for Plus users. Pro $100 and Pro $200 remain exempt for the upcoming months.

What happens when I hit the Codex 5-hour limit?

You cannot start new Codex turns until the window rolls over (after five hours of usage activity). You can wait, purchase credits, use an available reset, or upgrade your plan.

Why can one Codex task consume more usage than another?

Consumption depends on model, repository context size, reasoning, tool usage, and output length. Larger tasks inherently consume more tokens.

How do I check my Codex usage?

In the Codex CLI, run /status. In ChatGPT, open Codex settings and view the Usage panel.

Does hitting the Codex limit affect regular ChatGPT?

 No. Codex and regular ChatGPT chat have separate usage models. ChatGPT Work shares the Codex allowance.

Can I get more Codex usage without upgrading?

Yes. You can wait for a reset, use a banked reset if available, or purchase additional credits.

What is the difference between the 5-hour and weekly limits 

The 5-hour window is a rolling constraint on short-term usage. The weekly limit is a separate cap on total usage across a week. You can hit one while having capacity under the other.

Do Pro users have the 5-hour limit?

Pro $100 and Pro $200 users will not have the 5-hour limit enabled for the upcoming months, though they remain subject to weekly limits.

Can I use cached input to reduce consumption?

Yes. Cached input costs roughly a tenth of fresh input, making it the most effective way to extend your allowance.

Is the 5-hour limit permanent?

OpenAI has not stated whether this is permanent. The company has adjusted limits multiple times in 2026. Treat current documentation as authoritative.