Codex Brings Back the 5-Hour Limit for Plus Users, While Pro Plans Stay Exempt

What Happened

OpenAI is bringing back the 5-hour usage limit for Codex on ChatGPT Plus, reversing the temporary removal of the restriction earlier this summer.

The change applies to Plus users across Codex and related agentic usage. However, Pro subscribers on both the $100 and $200 tiers will remain exempt from the 5-hour limit for the coming months.

According to an explanation from Tibo on the Codex team, OpenAI has two main reasons for restoring the restriction on Plus accounts:

• Balancing compute demand: A 5-hour usage window helps OpenAI smooth spikes in Codex workloads and distribute available compute capacity more evenly.
• Protecting weekly allowances: Plus users have relatively limited included usage, and unrestricted short-term consumption can cause someone to accidentally burn through most or all of their weekly allowance in a single intensive session.

The second point is particularly important because the 5-hour limit should not be confused with a five-hour working-time restriction. It is effectively a rolling usage bucket, and complex coding sessions can consume that allowance much faster than five hours of real time.

Large repositories, long contexts, higher reasoning settings, repeated tool calls, autonomous testing, and more demanding models can all increase consumption.

OpenAI had temporarily removed the 5-hour restriction from several Codex plans in July while evaluating the impact on usage patterns and infrastructure demand. The latest change represents a return to short-term throttling for Plus while maintaining a different policy for higher-priced Pro subscriptions.

Importantly, Pro users are not receiving unlimited Codex usage. The absence of the 5-hour restriction only means that their included allowance can be consumed more freely during intensive sessions. Weekly usage limits still apply.

Why It Matters

The change creates a clearer distinction between ChatGPT Plus and ChatGPT Pro for serious Codex users.

Plus is increasingly positioned for developers who use Codex regularly but at a relatively steady pace. The combination of a 5-hour window and a weekly allowance prevents unusually heavy sessions from consuming a disproportionate amount of compute.

Pro users, meanwhile, can currently concentrate significantly more of their weekly usage into a short period without being stopped by the additional 5-hour bucket.

That difference can be substantial for software development workflows, where usage is rarely evenly distributed. A developer might barely use Codex for several days and then spend an entire day building a feature, refactoring a repository, running tests, reviewing code, and fixing issues with multiple agent sessions.

Under a strict 5-hour window, the user can encounter a temporary limit even when weekly capacity remains available. Without that restriction, a Pro subscriber can continue consuming the remaining weekly allowance during the same development session.

The current structure therefore looks roughly like this:

• ChatGPT Plus: weekly allowance + 5-hour usage restriction
• Pro $100: larger weekly allowance + no 5-hour restriction for now
• Pro $200: substantially larger weekly allowance + no 5-hour restriction for now

This makes the absence of the 5-hour limit a meaningful Pro benefit in addition to the higher usage multipliers associated with those subscriptions.

The policy also reveals how OpenAI is managing the economics of agentic coding. Codex workloads can involve substantially more compute than a traditional chatbot interaction because an agent may repeatedly inspect files, reason over a large repository, execute commands, run tests, analyze failures, and continue working autonomously.

A short-term rate limit gives OpenAI another mechanism for controlling peak demand without reducing the total weekly allowance available to a subscriber.

For Plus users, the limit also acts as a form of consumption protection. Without it, a single aggressive coding session could potentially consume enough of the weekly allowance to leave little included capacity for the rest of the week.

What's Next

The most important detail is that the Pro exemption is not described as permanent.

OpenAI's current position is that the 5-hour restriction will remain disabled for the $100 and $200 Pro tiers for the coming months. That wording leaves room for the company to adjust the policy again as Codex usage, infrastructure capacity, and subscription economics evolve.

Developers should therefore treat the current Pro behavior as a product policy rather than a permanent guarantee.

The broader direction is also becoming clearer. OpenAI appears to be separating Codex access across three dimensions:

• Total included usage over a longer period
• Short-term consumption limits that control burst usage
• Additional paid usage after included allowances are exhausted

That model allows OpenAI to support both casual coding and highly intensive agent workflows without giving every subscription tier identical access to peak compute capacity.

For developers deciding between plans, the practical question is no longer simply how many Codex requests a subscription includes. How quickly that allowance can be consumed is becoming equally important.

Users who occasionally ask Codex to modify files, debug code, or implement isolated features may see little reason to move beyond Plus.

Developers who routinely run long autonomous sessions, work with large repositories, use higher reasoning settings, or concentrate most of their coding into a few intensive sessions may find the Pro tiers substantially more useful because the 5-hour throttle is currently absent.

The key limitation remains the same for every heavy user: no 5-hour limit does not mean unlimited Codex. Pro subscribers can still exhaust their weekly allowance if usage is sufficiently intensive.

OpenAI is likely to keep refining these limits as it gathers more data about how developers use increasingly autonomous coding agents. For now, the restored Plus restriction establishes a much clearer boundary between everyday Codex access and the burst capacity reserved for Pro subscribers.