Codex Usage Limits: How to Check Them on Mac
Short answer: Codex limits come from your ChatGPT plan and reset on a five-hour window, with weekly limits that may also apply. Plus gets roughly 15 to 160 local messages per five hours on GPT-6.1 Sol, depending on task size; Pro currently has no five-hour limit. Check what’s left with /status in the Codex CLI or at chatgpt.com/codex/settings/usage.
Checked 14 October 2026 against OpenAI’s Codex pricing page. CLI behaviour observed in Codex CLI 0.160.0. OpenAI changes models and limits often, so treat the numbers as a snapshot.
TL;DR
- Limits depend on plan and model. Plus and standard Business share the same five-hour ranges; Pro has no five-hour limit for now.
- Two windows: a 5-hour window and, where it applies, a weekly one. The CLI logs both.
- Check it:
/statusin the CLI, or the usage dashboard atchatgpt.com/codex/settings/usage. - At the limit: the current turn finishes, then wait, buy credits or switch to an API key.
- Shared allowance: CLI, IDE extension, desktop app and cloud tasks all count against the same plan.
- Keep it visible in the Mac menu bar with a usage tracker such as FavTray (free).
What are the Codex usage limits for each plan?
OpenAI lists plans and limits on its Codex pricing page. On 14 October 2026:
| Plan | Price | Codex access | Limits |
|---|---|---|---|
| Free | $0 | Desktop app, GPT-6 Luna at Standard speed, subject to rollout | Not published |
| Go | $8/month | Same as Free | Not published |
| Plus | $20/month | Web, CLI, IDE extension and iOS; GPT-6.1 Sol and GPT-6 Luna | Per-5-hour ranges below; weekly limits may apply; extend with credits |
| Pro | $100, $200 or $500/month | Everything in Plus; Ultrafast mode on the $500 tier | No five-hour limit currently |
| Business | $20/user/month yearly (2+ users), $25 monthly | Workspace, SSO | Same ranges as Plus; workspace credits |
| Enterprise and Edu | Contact sales | Workspace | Flexible pricing has no fixed rate limits |
| API key | API rates | Local only, no cloud features | Billed per token |
Local messages per five hours, Plus and standard Business:
| Model | Messages per 5 hours |
|---|---|
| GPT-6.1 Sol | 15 to 160 |
| GPT-6 Sol | 15 to 150 |
| GPT-6 Astra | 5 to 45 |
| GPT-6 Luna | 350 to 3,000 |
Why a range? A message that reads a large repo and runs for ten minutes costs far more than a one-line question. OpenAI calls these estimates, not fixed limits.
Two multipliers eat into the allowance faster:
- Fast speed uses 2.5 times the included usage of Standard on the same model.
- Ultrafast (GPT-6 Astra and GPT-6.1 Sol) uses 8 times.
Also note that GPT-5.5 retires from Codex with ChatGPT sign-in on 14 October 2026 at 19:00 UTC; API-key sessions aren’t affected. If you had it pinned in config.toml, pick another model.
How do the five-hour and weekly windows work?
Codex tracks usage as a percentage of each window, not as a message count you can see. In Codex CLI 0.160.0, every session log records two windows:
- a primary window of 300 minutes (five hours), and
- a secondary window of 10,080 minutes (one week),
each with used_percent and a resets_at time. Your five-hour window and your weekly window run at the same time. Hitting either one stops new work until it resets.
Everything shares the allowance: local messages in the CLI and IDE extension, the desktop app, and cloud tasks. Cloud tasks can use more of it than the same work done locally. Code reviews that Codex runs through GitHub count separately; reviews you run locally count as normal usage.
How do I check my Codex usage?
In the Codex CLI: /status
- Start or open a Codex session with
codex. - Type
/statusand press Return. - If it says the data isn’t available yet, send one message and run
/statusagain shortly.
It shows the session details, context use and your rate limits: how much of the five-hour and weekly windows is used, and when each resets. In the CLI, /usage shows your daily, weekly or cumulative token activity. In the desktop app, /status works from the composer too, and /usage opens usage and billing settings when you’re signed in with ChatGPT (desktop slash commands).
In the browser: the usage dashboard
Open chatgpt.com/codex/settings/usage while signed in to the same ChatGPT account. OpenAI points here for current limits and reset times, and it’s where you’d buy credits on Plus or Pro.
From the session logs (read-only)
The CLI writes each session to ~/.codex/sessions/ (or $CODEX_HOME/sessions/). To read the last limits it saw, without starting Codex:
f=$(ls -t ~/.codex/sessions/*/*/*/*.jsonl | head -1)
grep -o '"rate_limits":{[^}]*}[^}]*}[^}]*}' "$f" | tail -1
You’ll get something like "rate_limits":{"limit_id":"codex",...,"primary":{"used_percent":12.0,"window_minutes":300,"resets_at":1791999735},"secondary":{...}. resets_at is a Unix timestamp; date -r 1791999735 turns it into a local time.
This only reads a file. It’s a snapshot from your last message, not a live figure, and the log format is internal to Codex, so it can change between versions.
With an API key
If you signed in with an API key (printenv OPENAI_API_KEY | codex login --with-api-key), there’s no plan limit to check. You pay per token, and spend shows in your OpenAI API dashboard. Our OpenAI API cost calculator helps estimate it.
What happens when you hit the Codex limit?
- The current turn finishes. OpenAI says the agent can keep working on a turn that’s already running, subject to fair use.
- New work waits until the window resets.
/statusor the dashboard tells you when. - Or extend it:
- Plus and Pro: buy ChatGPT credits from the usage dashboard.
- Business, Edu, Enterprise (flexible pricing): use workspace credits.
- Anyone: sign in with an API key for extra local work at standard API rates.
Credits are priced per million tokens and vary by model. OpenAI’s page gives GPT-6.1 Sol at 50 credits per million input tokens and 250 per million output tokens, and GPT-6 Luna at 2.5 and 12.5.
How do you make your Codex limit last longer?
- Pick the model for the job. GPT-6 Luna’s range is far higher than GPT-6.1 Sol’s. Use the bigger model for planning and hard bugs, and the smaller one for edits and tests.
- Stay on Standard speed unless you need the answer now. Fast costs 2.5 times as much allowance.
- Prefer local over cloud tasks for routine work.
- Keep context small. Point Codex at the files that matter instead of letting it read the whole repo every turn.
- Check before a long task. Starting a big refactor at 90% of your five-hour window is how you end up stuck halfway.
Several ChatGPT accounts, such as personal and work? Each has its own limits, and nothing is pooled. Our guide to switching Codex CLI accounts covers CODEX_HOME and logins, within OpenAI’s terms.
How do you see Codex limits in the Mac menu bar?
/status answers when you ask. A menu bar app shows the numbers all day, including when no session is open. We make FavTray, so weigh our take accordingly.
FavTray AI Usage is free on every plan, with full history. For Codex it shows:
- the session and weekly limits with reset times, a pace marker and a forecast of when you’ll hit the limit at your current rate;
- warnings at 80% and 95%, and a notification when a limit runs out and when it’s back;
- spend by model and by day, worked out from the Codex logs on your Mac;
- Claude, Cursor, Gemini, Copilot and other providers on the same screen, if you use more than one tool.
It talks to OpenAI with the sign-in you already have, and no FavTray server sees your usage. Saving several ChatGPT logins and switching Codex between them from the menu bar is part of Pro, which also suggests a saved account with room when the active one is near its limit. It never switches for you.
Other options: Claude Usage Tracker (free, open source) added Codex in v3.3.0, and SessionWatcher (paid) covers Codex in its Bundle and Pro tiers. We compare them, along with the log-based CLI ccusage, in ccusage alternatives for Mac. If you also run Claude Code, how to see Claude Code usage limits on Mac covers the Anthropic side, and OpenAI Codex vs Claude Code compares the two agents.
Quick reference
| You want to… | Do this |
|---|---|
| See what’s left right now | /status in a Codex session |
| See reset times and buy credits | chatgpt.com/codex/settings/usage |
| Read the last known limits without starting Codex | grep the latest session log (read-only) |
| Keep limits visible all day | A menu bar tracker such as FavTray AI Usage |
| Keep working after the limit | Credits, or an API key at API rates |
FavTray’s AI Usage tracker is free; the 7-day Pro trial needs no card, and Pro is $4.99 once (pricing).
Sources (checked 14 October 2026): Codex pricing and usage limits, Codex CLI slash commands, desktop app slash commands, Codex authentication. Log fields observed in Codex CLI 0.160.0 and may change.
Frequently Asked Questions
What are the Codex usage limits on ChatGPT Plus?
OpenAI gives Plus and standard Business seats a range of local messages per five-hour window that depends on the model: roughly 15 to 160 for GPT-6.1 Sol, 5 to 45 for GPT-6 Astra and 350 to 3,000 for GPT-6 Luna, as listed on 14 October 2026. Weekly limits may also apply. OpenAI calls these estimates, because longer tasks use more.
How do I check my Codex usage?
Run /status in an active Codex CLI session to see your remaining limits, or open the usage dashboard at chatgpt.com/codex/settings/usage for current limits and reset times. Both use your ChatGPT sign-in. If you sign in to Codex with an API key, check usage in your OpenAI API dashboard instead, since you're billed per token.
Does Codex have a weekly limit?
It can. OpenAI's pricing page says weekly limits may apply on top of the five-hour window. Recent Codex CLI session logs report two windows, one of 300 minutes and one of 10,080 minutes (a week), each with a percentage used and a reset time. The usage dashboard shows the reset time for each.
What happens when I hit the Codex usage limit?
If you hit the limit in the middle of a turn, Codex finishes that turn, subject to fair use. After that you can wait for the window to reset, buy ChatGPT credits on Plus or Pro, use workspace credits on Business, Edu or Enterprise, or sign in with an API key and pay standard API rates for extra local work.
Does ChatGPT Pro have Codex usage limits?
OpenAI's pricing page says Pro plans currently have no five-hour limit. Pro comes in $100, $200 and $500 monthly tiers, and Ultrafast mode is available on the $500 tier. Fair use still applies, and OpenAI can change limits, so check the usage dashboard for what applies to your account.
Do Codex CLI and the Codex app share the same limits?
Yes. OpenAI says local messages and cloud chats share your plan's usage allowance, and Codex in the CLI, IDE extension, desktop app and web all draw from it. Cloud tasks can use more of the allowance than local messages. Code reviews that Codex runs through GitHub are counted separately.