OpenAI Codex Usage Limits by Plan and Model

OpenAI Codex usage limits vary by subscription tier and underlying model, calculated as estimated ranges of local messages allowed within rolling five-hour intervals.

At Layer3Labs, we build custom agents and workflow automations for small and mid-sized businesses, and a Codex cap that trips mid-run stops every step wired behind it.

OpenAI does not enforce static message numbers across all accounts. Plan allowances are shared across Codex, ChatGPT Work, and workspace agents, with five-hour and weekly reset windows determining when your allowance refreshes.

Codex Usage Limits by Plan and Model

OpenAI measures Codex allowances using estimated ranges of local messages allowed within rolling five-hour intervals rather than rigid numeric caps. OpenAI can change these ranges without notice, so confirm your current limits and reset times on the Codex pricing page or in your usage dashboard.

According to the official OpenAI documentation, Codex access is included with Free, Go ($8 monthly), Plus ($20 monthly), Pro (from $100 monthly), Business ($20 per user monthly on annual billing with a 2-user minimum, or $25 monthly), and Enterprise agreements. The Free and Go tiers grant access to GPT-6 Luna at Standard speed in the desktop app, subject to rollout schedules, but OpenAI publishes no specific five-hour message estimates for Free or Go.

On paid tiers, the volume of prompts you can submit every five hours depends heavily on the model you select in the desktop app, Integrated Development Environment (IDE) extension, or Command Line Interface (CLI). High-reasoning flagship models consume your allowance much faster than lightweight helper models.

• GPT-6 Astra: 5 to 45 messages on Plus and Standard Business; 25 to 225 on Pro 5x and the $100 Business tier; 100 to 900 on Pro 20x
• GPT-6 Sol: 15 to 150 messages on Plus and Standard Business; 70 to 700 on Pro 5x; 300 to 3,000 on Pro 20x
• GPT-6 Luna: 350 to 3,000 messages on Plus and Standard Business; 1,750 to 14,000 on Pro 5x; 7,000 to 56,000 on Pro 20x
• GPT-5.6 Sol: 10 to 100 messages on Plus and Standard Business; 50 to 500 on Pro 5x; 200 to 2,000 on Pro 20x
• GPT-5.6 Terra: 25 to 200 messages on Plus and Standard Business; 125 to 1,000 on Pro 5x; 500 to 4,000 on Pro 20x
• GPT-5.6 Luna: 250 to 2,000 messages on Plus and Standard Business; 1,250 to 10,000 on Pro 5x; 5,000 to 40,000 on Pro 20x
• GPT-5.5: 15 to 80 messages on Plus and Standard Business; 75 to 400 on Pro 5x; 300 to 1,600 on Pro 20x
• GPT-5.4: 20 to 100 messages on Plus and Standard Business; 100 to 500 on Pro 5x; 400 to 2,000 on Pro 20x (removed from Codex with ChatGPT sign-in on August 31, 2026)
• GPT-5.4 mini: 60 to 350 messages on Plus and Standard Business; 300 to 1,750 on Pro 5x; 1,200 to 7,000 on Pro 20x (removed from Codex with ChatGPT sign-in on August 31, 2026)

Mechanics of Rolling Five-Hour and Weekly Reset Windows

OpenAI limits Codex over a rolling five-hour window and, depending on your plan, a weekly window. As detailed in the OpenAI Astra usage guide, where both apply, you need allowance left in both to keep working.

A five-hour window does not follow fixed clock intervals. The timer begins the moment you send your first message in Codex or ChatGPT Work after your previous window expires. If you exhaust your allowance in two hours, you must wait the remaining three hours for that specific window to reset, even if your weekly allotment has ample capacity.

If you hit plan boundaries while Codex is actively generating code, OpenAI permits the agent to complete that active turn under standard fair-use provisions. Once the turn completes, subsequent prompts remain blocked until the reset timestamp arrives or you add credits. OpenAI Support does not manually reset account limits upon request.

• Rolling five-hour reset: initiates immediately upon sending the first prompt after the prior window ends
• Weekly allowance window: controls total aggregate usage across seven continuous days
• Dual-requirement rule: both windows must retain available capacity for prompts to process
• Turn preservation: active code execution continues to completion during an exhausted turn under fair-use safeguards
• Banked resets: activating a banked reset refreshes both the five-hour and weekly windows while establishing a new weekly reset date

Monitoring and Verification of Account Consumption

Developers can verify active Codex consumption using the Command Line Interface status command or the account web dashboard. Running the /status command directly inside an active terminal session displays your current exhaustion state and remaining allowances.

You can inspect exact quotas by visiting the OpenAI usage settings in your browser. This screen lists whether your five-hour or weekly ceiling has been hit, provides active credit balances, and shows the exact time your limits will reset.

For developers using the desktop application, navigate to Settings and select Usage. Business and Enterprise users can monitor Codex and Work usage limits in Codex settings under Usage, per the OpenAI rate card.

Methods for Managing Codex Usage Limits and Overage

Developers who reach their included message ceiling can purchase ChatGPT credits, upgrade their account tier, or connect an Application Programming Interface (API) key.

Personal Plus users paying $20 monthly can upgrade to the Pro $100 plan to unlock five times the base Plus usage allowance. The Pro $200 plan provides twenty times the Plus limit, but the OpenAI Pro tiers article confirms that OpenAI temporarily paused new sign-ups and upgrades to the Pro $200 tier on September 10, 2026. Existing Pro $200 subscribers retain their access, while new subscribers must choose the Pro $100 tier.

Users on Plus and Pro can buy flexible credits directly inside account settings without changing subscription tiers, as outlined in the OpenAI credit guide. Included plan allowances burn first; purchased credits activate only after plan quotas exhaust. These credits expire twelve months from purchase, cannot be transferred between accounts, and remain completely distinct from OpenAI API credits.

Adding a personal API key routes your requests through standard pay-as-you-go developer rates. As shown on the OpenAI API pricing documentation, standard token charges apply: GPT-6 Sol costs $2.00 per million input tokens and $10.00 per million output tokens on short contexts, while GPT-6 Astra costs $10.00 input and $50.00 output. The dedicated gpt-5.3-codex model costs $1.75 input and $14.00 output. Using an API key disables cloud features such as automatic GitHub code review.

Consumption Multipliers and Lifecycle Timelines

Operating Codex in specialized execution modes accelerates allowance depletion through higher credit and message multipliers. Fast mode consumes 2.5 times the standard credit rate for GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna.

Cloud chats on ChatGPT plans use GPT-5.6 Sol and may use more of your allowance than local messages. Image generations also use your included limits roughly three to five times faster on average.

According to the OpenAI Business rate card, average developer consumption across organizations ranges from $100 to $200 per developer monthly when factoring in model selection, concurrency, and automated triggers. GPT-5.6 Sol promotional credit rates run through at least November 21, 2026. Older models face mandatory retirement: GPT-5.5 retires from Codex across all plans on October 14, 2026, while GPT-5.4 and GPT-5.4 mini access ended on August 31, 2026 for accounts signing in via ChatGPT.

Workflow Allocation and Strategic Recommendations

Picking a lighter model for routine work stretches your allowance, because GPT-6 Luna's Plus range (350 to 3,000 messages) far exceeds GPT-6 Astra's (5 to 45). Direct everyday syntax generation and basic file refactoring to GPT-6 Luna or GPT-6 Sol, reserving GPT-6 Astra strictly for multi-file system architecture and complex debugging sessions.

In the agent pipelines we build for clients, a loop that polls branches on GPT-6 Astra is the pattern that hits the five-hour cap first. Teams running high-throughput autonomous loops avoid sudden blocks by provisioning dedicated API keys rather than relying solely on shared seat allowances.

Codex seat-based tiers do not fit teams requiring completely uncapped autonomous agent swarms that run continuous overnight code generation without token accounting. Engineering shops with that workload profile should route tasks through the raw API or review our OpenAI Codex versus Claude Code breakdown and Claude Code pricing analysis to compare alternative consumption models.

Our recommendation would change if OpenAI removes the rolling five-hour window entirely or reopens new subscriptions to the Pro $200 twenty-times multiplier tier. For plan prices, see our OpenAI Codex pricing guide, and check your remaining allowance in the Codex usage dashboard before running multi-file edits.

How to use Codex

You do not host Codex yourself — you use it through a tool, so "getting started" really means choosing the right one.

The fastest way to put Codex to work day to day is inside an AI IDE, and Cursor is the most popular — it supports it directly, so you can be working in minutes. The maker's own option is Codex for Codex, if you want the native experience. Prefer a different editor? Windsurf, Zed, and GitHub Copilot drive these models too.