OpenAI Codex Usage Limit and Reset Time: Free vs Paid Tier Guide

OpenAI Codex (the agentic CLI coding tool, not the legacy code-completion API) runs on a usage-limited model. Free tier users get approximately 10 tasks per day, with the limit resetting every 24 hours from your first task. ChatGPT Pro subscribers ($200/month) get significantly higher — effectively uncapped — task limits. If you're hitting walls mid-session, here's exactly what the limits look like and when they roll over.

what are the codex usage limits by plan?

Codex launched in preview in 2025 as an open-source CLI agent built on the o3 and o4-mini model family. OpenAI has kept the exact task caps deliberately vague, but here's what's known from OpenAI's Codex documentation and community reports:

Plan	Daily Task Limit	Model Access	Reset Window	Cost
Free (ChatGPT)	~10 tasks/day	o4-mini (limited)	24 hours rolling	$0
ChatGPT Plus	Higher cap (varies)	o4-mini + o3	24 hours rolling	$20/month
ChatGPT Pro	Near-unlimited	o3 (full access)	Monthly billing cycle	$200/month
API (pay-as-you-go)	Token-based, no hard cap	codex-mini-latest	N/A (metered)	Per token

The free tier runs on codex-mini-latest, a fine-tuned variant of o4-mini optimized for code. According to OpenAI's Codex pricing page, API access is priced at $1.50 per million input tokens and $6.00 per million output tokens — with an 80% discount for cached input tokens.

when does codex usage reset?

For free and Plus users, the Codex task counter resets on a rolling 24-hour window — not at midnight UTC. The clock starts from when you ran your first task. So if you hit your limit at 3pm Tuesday, you'll get fresh quota around 3pm Wednesday.

Pro subscribers operate on a monthly billing cycle. Their limits (where applicable) reset on the same date each month that their subscription renewed.

API users have no reset to worry about. Usage is metered per token, billed to your OpenAI account, and subject only to your account-level rate limits (requests per minute, tokens per minute).

codex rate limits: requests per minute and tokens per minute

Beyond daily task caps, Codex also enforces API-level rate limits for developers using the REST API directly. These vary by usage tier (not subscription plan — OpenAI assigns API tiers based on cumulative spend):

API Tier	Requests per Minute	Tokens per Minute	Minimum Spend to Unlock
Tier 1	500 RPM	200,000 TPM	$5 paid
Tier 2	1,000 RPM	2,000,000 TPM	$50 paid
Tier 3	5,000 RPM	4,000,000 TPM	$100 paid
Tier 4	10,000 RPM	10,000,000 TPM	$250 paid
Tier 5	30,000 RPM	Unlimited	$1,000 paid

If you're a solo developer on a new API account, you'll start at Tier 1 by default. Rate limit errors (429 Too Many Requests) at this tier are common during heavy automated workflows.

what happens when you hit the codex usage limit?

When you exceed the free tier daily cap in the Codex CLI, you'll see an error along the lines of:

Error: You have exceeded your daily usage limit. Please try again later or upgrade your plan.

Via the API, hitting a rate limit returns a standard 429 status with a Retry-After header indicating how many seconds to wait. Most OpenAI SDK clients handle this automatically with exponential backoff.

The cleanest fix for free-tier CLI users is to upgrade to ChatGPT Plus or Pro. For API developers, either reduce request frequency or apply to increase your tier by spending toward the next threshold.

codex free tier: is it enough for real work?

Ten tasks per day sounds limited, but "task" in Codex terms is a full agentic run — the tool can spin up a sandbox, write files, run tests, and iterate autonomously. For light exploration or learning, the free tier is workable.

For production use or active development, you'll exhaust 10 tasks before lunch. The upgrade calculus is straightforward: Plus at $20/month unlocks meaningfully more headroom; Pro at $200/month is aimed at teams or developers where Codex is a primary workflow tool.

Compare this to how Claude Code handles limits — it uses a 5-hour rolling window plus a weekly cap, which feels different in practice (you can burst hard in short windows). Codex's daily task model is more predictable but less flexible for power users.

how to check your codex usage

There's no built-in /usage command in Codex CLI the way Claude Code has one. Your options:

• OpenAI platform dashboard: Go to platform.openai.com/usage to see token consumption across all API calls by day, model, and project.
• ChatGPT settings: For subscription-based Codex usage (not API), check your ChatGPT account settings for a usage summary.
• API response headers: Each API response includes x-ratelimit-remaining-requests and x-ratelimit-remaining-tokens headers you can log programmatically.

There's no menubar-style live tracker for Codex the way Usagebar provides for Claude Code — though if you're splitting time between both tools, Usagebar's real-time Claude Code window tracking can help you stay aware of where you stand on the Anthropic side.

codex vs claude code: usage model comparison

If you're evaluating both tools, the usage models are structurally different. Claude Code uses subscription-based access through Claude Pro or Max plans with soft usage windows. Codex uses either a task-count cap (free/Plus) or pure token metering (API). See our full Codex vs Claude Code comparison for a deeper breakdown.

For developers hitting Claude Code limits frequently, the Claude Code reset timing follows a different pattern — worth understanding if you context-switch between both tools during the day.

key takeaways

• Free Codex CLI: ~10 tasks/day, resets on a 24-hour rolling window from first use
• ChatGPT Plus/Pro: higher task caps, Pro is near-unlimited
• Codex API: no daily cap, metered by token, rate limits by API tier (500-30,000 RPM)
• API pricing: $1.50/M input tokens, $6.00/M output tokens (80% cache discount)
• Check usage at platform.openai.com/usage or via response headers
• No built-in live usage tracker exists for Codex; plan accordingly

sources

• https://developers.openai.com/codex
• https://developers.openai.com/codex/pricing
• https://usagebar.com/blog/codex-vs-claude-code
• https://usagebar.com/blog/when-does-claude-code-usage-reset
• https://usagebar.com/blog/claude-code-weekly-limit-vs-5-hour-lockout
• https://usagebar.com/blog/openai-codex-pricing