Pricing

Codex Usage Limits (2026): By Plan, Reset Window and How to Check

Codex usage limits by ChatGPT plan, when they reset, where to check what you have left, and your options at the cap. Only OpenAI-published numbers.

By Engineer Foodie Desk Oct 26, 2026 ✓ Sources checked Oct 26, 2026

8 min read

Limits and reset rules checked Oct 26, 2026 on OpenAI’s Codex pricing page and OpenAI’s Codex help article; other dates are under Sources. No affiliate links on this page (how we work).

Codex limits run on two clocks: a five-hour window and, on top of it, weekly limits that “may also apply”. OpenAI publishes message estimates for exactly one tier of individual user, Plus (and Standard Business, which matches it). Pro “currently” has no five-hour limit. Free and Go get no numbers at all. And the only reset times that count are the ones on your own usage dashboard, or /status in the CLI (verified Oct 26, 2026).

We haven’t tested these limits ourselves; every number below is OpenAI’s, with its own caveats left in.

Quick answer

Reset window: limits are counted “per five-hour period”, and “Weekly limits may also apply”. OpenAI doesn’t publish the weekly allowance or a fixed reset hour.
Plus: 15–160 local messages per five hours on GPT-6.1 Sol, 350–3,000 on GPT-6 Luna (OpenAI estimates, “not fixed message limits”).
Pro: “Pro plans currently have no five-hour limit.”
Free and Go: “Limited”, with no number published.
Check usage: your usage dashboard (Settings > Usage), or type /status in an active Codex CLI session.
At the cap: finish the current turn, then buy credits, use a banked reset if you have one, upgrade, or wait.

Codex limits by plan

Plan	Limit	Reset
Free	Not published	Not published
Go	Not published	Not published
Plus	See estimates below	5 hours + weekly
Pro	No 5-hour limit	Weekly not published
Business	Same as Plus	5 hours + weekly
Enterprise / Edu	Credits or Plus-level	Not published

Source: Codex pricing, verified Oct 26, 2026. Enterprise and Edu with flexible pricing have “no fixed rate limits. Usage scales with credits”; without flexible pricing they get “the same per-seat usage limits as Plus for most features”.

OpenAI’s estimates of local messages per five hours on Plus and Standard Business (verified Oct 26, 2026):

Model	Plus	Business
GPT-6 Astra	5–45	5–45
GPT-6.1 Sol	15–160	15–160
GPT-6 Sol	15–150	15–150
GPT-6 Luna	350–3,000	350–3,000

Why the ranges are so wide: “Model choice, context, reasoning, tool use, retrieval, and caching all affect usage, so prompt length alone isn’t a reliable estimate.” And two things shrink them further:

Cloud tasks cost more. “Cloud tasks may use more of your allowance than local messages.” Local messages and cloud chats “share your plan’s usage allowance.”
Speed costs usage. On included subscription usage, Fast mode burns limits at 2.5x and Ultrafast (GPT-6 Astra and GPT-6.1 Sol, Pro $500 only) at 8x. Image generation uses included limits “~3-5x faster on average”.

Codex also shares its pool. “ChatGPT Work and Codex share usage”, and the limits currently include ChatGPT for Excel on Plus and Pro. Your coding budget and your spreadsheet budget are the same budget.

Free and Go: the ChatGPT pricing grid says Codex is “Limited” on both, and the estimates table doesn’t include them (ChatGPT pricing, verified Oct 26, 2026).

A 5–45 range on GPT-6 Astra is a 9x spread, and the weekly cap has no number at all. OpenAI is honest that these aren’t fixed limits, but “check your dashboard” is doing a lot of work for a paid plan. A published weekly figure would save a lot of guessing.

Source: OpenAI Codex pricing, checked Oct 26, 2026

How the reset works

Here’s everything OpenAI publishes about resets, in its own words (all from the Codex pricing page and the help article, verified Oct 26, 2026):

The windows: estimates are per “five-hour period”, and “Weekly limits may also apply.” OpenAI doesn’t say whether the five-hour window starts with your first message or rolls continuously. That’s not published.
Your reset time lives on your account: “check your usage dashboard for current limits and reset times.”
A new chat doesn’t help: “Starting a new chat does not reset your usage limits.”
Support won’t reset you: “OpenAI Support does not reset ChatGPT or Codex usage limits.” If you think usage was miscounted, or access didn’t return after the stated reset time, Support will investigate.
Banked resets exist, sometimes. On some Plus and Pro referral promotions, “a successful referral can add a banked Codex rate-limit reset.” You use it from the profile menu’s usage summary (“1 reset available”). “Using a full banked reset refreshes your 5-hour and weekly Codex usage windows and changes your weekly reset date.”
One-time resets: OpenAI “may occasionally provide eligible users with a one-time reset”, but “Future resets are not guaranteed.”
Other ChatGPT limits are separate. Upload, image and voice banners in ChatGPT, like “resets in 720 hours”, “do not apply to Codex.”

Where to check your remaining usage

Three places, as OpenAI documents them (verified Oct 26, 2026):

Usage dashboard / Settings. “You can find your current limits in the usage dashboard.” The help article says to “open Settings or your usage dashboard. Check which allowance is exhausted, your credit balance, and any reset time shown.” After a banked reset, “Go to Settings > Usage to check your updated reset time.”
Codex CLI. “If you want to see your remaining limits during an active Codex CLI session, you can use /status.” The refreshed CLI also lists “Usage stats: Review usage information in the CLI.”
Credit balance. If you’ve bought credits, the balance is in “Settings > Usage in ChatGPT or Usage & Billing in the Codex app” (credits help article, verified Oct 24, 2026).

OpenAI’s own advice is to “Check the dashboard every week or two to understand your pace and remaining capacity.” For the IDE extension, OpenAI doesn’t document a separate usage view; it draws from the same Codex Local allowance as the CLI.

Hit the limit: wait, credits, or upgrade

First, the good news: “If you reach your usage limits during an active turn, the agent will be able to continue working on that turn, subject to fair use limits.” So the turn in progress gets finished (verified Oct 26, 2026). After that turn:

Wait for the reset time shown on your dashboard. Free, slow.
Buy 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.” Per-credit price isn’t published, so check current price at checkout. Details: Codex credits.
Use a banked reset, if a referral promotion gave you one.
Use an API key. “All users may also run extra local chats using an API key, with usage charged at standard API rates.”
Upgrade. Pro “currently” has no five-hour limit. Whether that’s worth $80 more a month is its own question: Codex Pro vs Plus.

OpenAI’s tips for stretching limits are specific and worth following: trim prompts, give only relevant files, shrink or nest your AGENTS.md, and disable MCP servers you don’t need, since “Every MCP server adds more context to your messages and uses more of your limit.” Picking GPT-6 Luna over GPT-6.1 Sol for routine work is the easiest lever on Plus: the estimate jumps from 15–160 to 350–3,000 messages.

Codex vs Claude Code limits

Claude Code on Pro and Max also uses a five-hour session limit plus a weekly limit, but Anthropic adds a fixed weekly reset per account that you can see in Settings > Usage (Claude Pro help, verified Oct 25, 2026). Codex Pro drops the five-hour window entirely for now, which Claude’s Max plans don’t; the full breakdown is in Claude Code usage limits, and the wider field in AI coding tools compared.

Here’s what I’d do

On Plus and hitting the wall: switch routine work to GPT-6 Luna and turn off Fast mode before spending a cent.

Still hitting it a few times a month: buy credits for those days.

Hitting it most days: that’s when Pro’s missing five-hour window earns its price. See Codex pricing for every tier.

Unsure where you stand: run /status mid-session, and look at Settings > Usage once a week.

OpenCode Go uses a similar five-hour and weekly pattern, measured in dollars; see OpenCode Go plan limits.

Your model choice changes how far a limit goes: GPT-6.1 Sol vs Terra shows the per-model message estimates side by side.

Wondering how Plus limits compare with Cursor’s? Cursor Pro vs Codex on Plus puts the two $20 plans side by side.

Not every Codex failure is a usage limit. If a run dies with “stream disconnected before completion”, see Codex “stream disconnected” error for the causes and fixes.

Seeing the exact message “You’ve hit your usage limit. Try again later.” in Codex? Codex’s usage-limit error explains which limit tripped and when it resets.

Comparing those limits with Claude’s? Codex Plus vs Claude Code Pro shows how each plan’s reset windows work.

FAQ

When do Codex usage limits reset? Codex counts usage per five-hour period, weekly limits may also apply, and your exact reset time is shown in your usage dashboard.

How do I check my Codex usage? Open Settings > Usage or your usage dashboard, or type /status in an active Codex CLI session.

What are Codex limits on Plus? OpenAI estimates 15–160 local messages per five hours on GPT-6.1 Sol and 350–3,000 on GPT-6 Luna, and weekly limits may also apply.

Does Codex Pro have usage limits? OpenAI says Pro plans currently have no five-hour limit, but it doesn’t publish Pro’s weekly allowance.

Does starting a new chat reset Codex limits? No, OpenAI says starting a new chat does not reset your usage limits.

Sources

OpenAI, Codex pricing (estimates, five-hour and weekly wording, cloud vs local, speed multipliers, /status, at-the-cap options, tips): https://developers.openai.com/codex/pricing (verified 2026-10-26)
OpenAI Help Center, Using Codex with your ChatGPT plan (Settings/dashboard, /status, banked and one-time resets, Support policy, new chats): https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan (verified 2026-10-26)
ChatGPT pricing (Codex Limited on Free and Go): https://chatgpt.com/pricing (verified 2026-10-26)
OpenAI Help Center, Using credits for flexible usage (credit balance location): https://help.openai.com/en/articles/12642688-using-credits-for-flexible-usage-in-chatgpt-personal-plans (verified 2026-10-24)
Anthropic, What is the Pro plan (five-hour session, weekly limit, fixed weekly reset): https://support.claude.com/en/articles/8325606-what-is-the-pro-plan (verified 2026-10-25)