OpenAI Codex Usage Limits Explained (2026)

OpenAI Codex Usage Limits Explained (2026)

OpenAI Codex is a powerful agentic coding tool, but it comes with usage limits that can be confusing. Whether you are using Codex through ChatGPT or the API, understanding these limits helps you plan your workflow and avoid hitting walls mid-project.

This guide explains every limit, how they work, and practical strategies to work within them or request increases.

Codex Limits by Plan

ChatGPT Plan Limits

If you access Codex through ChatGPT, your limits depend on your subscription tier.

Limit Type	Plus ($20/mo)	Pro ($200/mo)	Team ($25/user/mo)	Enterprise
Tasks per day	~25	~250	~100	Custom
Max files per task	50	200	100	Custom
Task timeout	10 minutes	30 minutes	15 minutes	Custom
Concurrent tasks	1	3	2	Custom
Repository size	Up to 500 MB	Up to 2 GB	Up to 1 GB	Custom
Model access	codex-mini	codex-mini + codex	codex-mini + codex	All

A few important notes:

• Task limits reset daily at midnight UTC.
• The ~25 task limit on Plus is approximate. OpenAI uses a "compute-based" limit, so simple tasks consume less of your quota than complex ones.
• Concurrent tasks refers to how many Codex tasks you can run simultaneously. Plus users can only run one at a time.

API Rate Limits

If you access Codex through the OpenAI API, you face different constraints.

Limit Type	Tier 1 (New)	Tier 2	Tier 3	Tier 4	Tier 5
RPM (requests/min)	60	100	300	800	2,000
TPM (tokens/min)	60,000	200,000	1,000,000	5,000,000	10,000,000
RPD (requests/day)	1,000	5,000	15,000	50,000	150,000
Monthly spend to qualify	$5	$50	$100	$250	$1,000

How API Tiers Work

OpenAI automatically assigns you a tier based on your cumulative spending and account age:

• Tier 1: New accounts with at least $5 in payments
• Tier 2: $50+ spent, account at least 7 days old
• Tier 3: $100+ spent, account at least 30 days old
• Tier 4: $250+ spent, account at least 30 days old
• Tier 5: $1,000+ spent, account at least 30 days old

You move up automatically. There is no application process for standard tier upgrades.

Understanding Token Limits

Every Codex interaction has token limits that affect how much context the model can process.

Model	Context Window	Max Output Tokens
codex-mini	200,000 tokens	16,384 tokens
codex	200,000 tokens	16,384 tokens

What affects token usage

• Your code files. Every file Codex reads counts as input tokens. A typical 200-line Python file is about 2,000-3,000 tokens.
• Your instruction. Your task description counts as input tokens.
• Generated code. The code Codex writes counts as output tokens.
• Internal reasoning. Codex's chain-of-thought reasoning (visible in the task log) counts as output tokens.

Estimating tokens for a task

Here is a rough guide for typical Codex tasks:

Task Complexity	Input Tokens	Output Tokens	Total Tokens
Simple (fix a bug in one file)	3,000-5,000	500-1,500	3,500-6,500
Medium (add a feature across 2-3 files)	10,000-30,000	2,000-5,000	12,000-35,000
Complex (refactor a module, write tests)	30,000-80,000	5,000-15,000	35,000-95,000

How to Check Your Current Usage

ChatGPT

1. Open ChatGPT and navigate to Settings > Subscription.
2. Your current usage and remaining tasks are shown under the Codex section.
3. You can also see a "Usage" indicator in the Codex interface itself.

API

Check your usage programmatically or through the dashboard.

import openai

# Check rate limit headers from any API response
response = openai.chat.completions.create(
    model="codex-mini",
    messages=[{"role": "user", "content": "Hello"}],
    max_tokens=10
)

# Rate limit info is in the response headers
# x-ratelimit-limit-requests: Your RPM limit
# x-ratelimit-remaining-requests: Remaining requests
# x-ratelimit-reset-requests: Time until reset

You can also check your usage dashboard at platform.openai.com/usage.

How to Increase Your Limits

Option 1: Upgrade Your ChatGPT Plan

The simplest way to increase Codex limits is to upgrade from Plus to Pro.

Metric	Plus	Pro (10x upgrade)
Daily tasks	~25	~250
Concurrent tasks	1	3
Task timeout	10 min	30 min
Price	$20/mo	$200/mo

Option 2: Move Up API Tiers

Increase your cumulative spend to automatically unlock higher tiers. The jump from Tier 1 to Tier 2 (just $50 total spend) gives you significant increases:

• RPM: 60 -> 100 (67% increase)
• TPM: 60,000 -> 200,000 (233% increase)
• RPD: 1,000 -> 5,000 (400% increase)

Option 3: Request a Custom Rate Limit Increase

For API users who need limits beyond Tier 5, OpenAI offers a rate limit increase request form.

1. Go to platform.openai.com/settings/organization/limits.
2. Click "Request rate limit increase."
3. Fill out the form with your use case, expected volume, and business details.
4. OpenAI typically responds within 2-7 business days.

Tips for approval:

• Explain your business use case clearly.
• Provide estimated daily/monthly token usage.
• Mention if you are building a product that integrates Codex.
• Higher-spending accounts get priority.

Option 4: Use Multiple API Keys

For organizations, you can create multiple API keys across separate projects. Each project gets its own rate limits. This is useful for teams where different projects have different usage patterns.

import openai

# Project A - main product
client_a = openai.OpenAI(
    api_key="sk-proj-A-...",
)

# Project B - internal tools
client_b = openai.OpenAI(
    api_key="sk-proj-B-...",
)

Strategies for Working Within Limits

1. Prioritize your tasks

Not every coding task needs Codex. Use it for complex, multi-step tasks and handle simple edits manually.

Good use of Codex tasks:

• "Refactor the payment module from callbacks to async/await and update all tests"
• "Add comprehensive error handling to every API endpoint in the routes/ directory"

Better done manually:

• Fixing a typo
• Changing a variable name
• Updating a version number

2. Optimize your repository structure

Create a .codexignore file to exclude unnecessary files from context:

# .codexignore - reduce token usage
node_modules/
dist/
build/
.git/
*.lock
*.map
coverage/
__pycache__/
.next/
vendor/

3. Write detailed instructions

Clear instructions help Codex complete tasks on the first attempt, avoiding the need to retry (which burns additional tasks from your quota).

Bad: "Fix the auth bug"

Good: "In src/auth/middleware.ts, the JWT verification on line 34 throws
an unhandled exception when the token is expired. Wrap it in a try-catch
that returns a 401 response with the message 'Token expired'. Add a unit
test in tests/auth.test.ts that verifies this behavior."

4. Use codex-mini for straightforward tasks

Reserve the full codex model for complex tasks that need deeper reasoning. The codex-mini model handles most single-file changes well and counts less against your compute-based quota.

5. Queue tasks efficiently

On the Pro plan, you can run 3 concurrent tasks. Queue your tasks strategically:

• Start a complex refactoring task first (takes longer).
• While it runs, submit two simpler tasks that complete quickly.
• Review results as they finish instead of waiting sequentially.

Common Error Messages and Fixes

Error	Meaning	Fix
Rate limit exceeded	Hit RPM or TPM limit	Wait and retry, or upgrade tier
Task limit reached	Daily task quota exhausted	Wait until midnight UTC, or upgrade plan
Context length exceeded	Too many tokens in context	Use .codexignore, reduce file scope
Task timed out	Task exceeded time limit	Break into smaller tasks
Concurrent task limit	Too many parallel tasks	Wait for current task to finish
Repository too large	Repo exceeds size limit	Exclude large files/directories

Frequently Asked Questions

Do unused tasks roll over to the next day? No. Daily task limits reset at midnight UTC and do not accumulate.

Does the "thinking" process count toward token limits? Yes. Codex's internal reasoning tokens count as output tokens for API billing. In ChatGPT, they count toward your compute-based task quota.

Can I see how many tasks I have remaining? In ChatGPT, yes -- the Codex interface shows your remaining quota. On the API, check the x-ratelimit-remaining-* response headers.

What happens if I exceed rate limits on the API? You receive a 429 (Too Many Requests) error. Implement exponential backoff in your code to handle this gracefully.

Are Codex limits separate from ChatGPT message limits? Yes. Codex has its own task quota that is separate from your ChatGPT conversation message limits.

Wrapping Up

OpenAI Codex limits vary significantly by plan -- from ~25 daily tasks on Plus to virtually unlimited on Enterprise. For most individual developers, the Plus plan is sufficient. If you consistently hit limits, upgrading to Pro or moving to the API with pay-per-token pricing gives you more flexibility.

If you are building AI-powered applications and need affordable media generation alongside your coding workflow, try Hypereal AI free -- 35 credits, no credit card required. The API is simple to integrate and works well with projects built using Codex or any other AI coding tool.