Codex Usage: How to Check Your Limits and Resets
To check your Codex usage, open Settings > Usage in ChatGPT or run /status inside an active Codex CLI session. These are the official account-level ways to see the allowances and reset times that apply to your account. OpenAI API usage at platform.openai.com/usage is separate from your ChatGPT subscription usage.
Verified October 1, 2026: chatgpt.com/settings/usage is the current web usage view we observed. Older Codex versions may still show the previous /codex/settings/usage link, which redirected to the ChatGPT home page in our signed-in check.
How to check Codex usage right now
Run inside any active Codex CLI session for instant turn-by-turn percentages:
- ✓Shows 5h window & weekly quota %
- ✓Shows exact reset time down to the minute
- ✕Requires entering a CLI session
Open the current ChatGPT Usage page in your browser without terminal access:
chatgpt.com/settings/usageThis is separate from metered API billing at platform.openai.com.
- ✓Shows 5h and weekly limits, reset times, credits & recent usage
- ✕Requires login; Analytics history may be delayed
Native macOS menu bar app, CLI, and local browser dashboard showing your real quota:
- ✓Menu bar companion, CLI & browser dashboard; refreshes every 60s
- ✓~30-second install, no separate account or API key
- Runs locally; sends no usage telemetry back to us
Included with Plus ($20) and Pro ($100/$200/$500). Usage windows depend on plan and account. Prompts pause when exhausted until the reset window ends.
Raw client-side token logs written to ~/.codex. Tools like ccusage read these locally, but local token counts do not reflect server quota or reset times.
Pay-as-you-go metered per million tokens for developer keys (OPENAI_API_KEY). Completely separate from ChatGPT subscriptions; exhausting prepaid balance returns HTTP 429.
What Counts as Codex Usage
Subscription quota tracks model inference calls executed by the Codex CLI, IDE extensions, or Codex Cloud background tasks under OpenAI's fair-use policies (Help Center #11369540). Local execution (such as test suites and git commands) runs on your own hardware without quota consumption, though tool output ingested back into conversation context enlarges prompt size. For image-related tasks, see how OpenAI meters image generation and credits.
- •Prompt turns in the CLI: Every user message and agent continuation sent to the model.
- •Codex Cloud tasks: Background agent tasks executed asynchronously in the cloud.
- •IDE edits & revisions: Full-file edits, multi-file changes, and inline suggestions.
- •Processed tool outputs: File contents, terminal outputs, and diffs read back into prompt context.
- •Local shell execution: Running commands like
npm testandgit commiton your machine. - •Direct disk inspection: Reading files, searching code with ripgrep, or navigating directories outside model turns.
- •Checking quota: Running
/statusor inspecting local CLI configuration. - •Metered API calls: Any request authenticated with an
OPENAI_API_KEYis billed separately on Platform balances.
How to Check Your Codex Usage (3 Methods)
Choose your checking method based on your immediate context:
Method 1: The /status Command in Codex CLI
Whenever you are inside an active Codex CLI session, type /status and press Enter. Codex immediately outputs your remaining 5-hour and weekly quota percentages alongside exact reset timestamps:
Note: The 5-hour and weekly percentages reflect your account subscription quota (see our deep dives on the 5-hour limit and weekly limit). They are distinct from session context memory, which the CLI footer reports separately as the share of the model's context window still available. To understand token compaction, read our Codex context window guide, or see our dedicated guide on how to check Codex usage in the CLI.
Method 2: ChatGPT Usage and Codex Analytics
Open chatgpt.com/settings/usage for the current web view. In our signed-in Plus check on October 1, 2026, the Overview tab showed the remaining 5-hour and weekly limits, reset countdowns, credits, usage-limit resets, and recent daily usage. The Analytics tab added usage history, historical 5-hour and weekly consumption, tool activity, chats, and messages.
OpenAI also still serves a separate Codex and Work Analytics page with quota balances plus breakdowns such as usage by surface, turns by model, plugin calls, and skills used. API usage remains separate at platform.openai.com/usage.
For a fuller comparison of the official web views, CLI output, and local monitoring, see our Codex usage dashboard and analytics guide.
Method 3: Local Codex Usage Companion & Dashboard
Instead of running /status manually, install Codex Usage, our free local tool. It connects directly to your local Codex authentication to monitor every reported quota limit, detect your most pressured window, and display countdown timers via a native macOS menu bar companion, terminal CLI, or local web dashboard (codex-usage --web) that refreshes every 60 seconds with zero external telemetry.
If you are looking for itemized token counts for individual sessions, tools like ccusage scan local session files stored in ~/.codex/sessions on your disk. These tools calculate total input and output tokens for your sessions, but they do not connect to OpenAI servers, cannot report remaining subscription quota percentages, and do not track 5-hour or weekly reset schedules.
Your Limits by Plan
Codex Quotas by Plan Tier
OpenAI publishes estimated local message ranges for Plus and Business Standard, sells Pro at $100/mo (5x), $200/mo (current allowance unquantified), and $500/mo (highest included Pro usage; Astra Ultrafast), and drops the 5-hour limit on Business Premium. It does not publish exact weekly allowances. Oldest fact verification: 27 Sep 2026.
| Plan | 5-Hour Window | Weekly Window | Token Limit | What Happens at the Limit | Source |
|---|---|---|---|---|---|
| Free ($0/mo) | Not published by OpenAI | Exact allowance not published by OpenAI | No universal subscription token cap published | Check reset time and options offered to your account | OpenAI Work and Codex pricing |
| Plus ($20/mo) | 5–3,000 (Codex Usage summary of OpenAI's model-specific estimates) local messages / 5h | Exact allowance not published by OpenAI | No universal subscription token cap published | Check reset time and options offered to your account | OpenAI Work and Codex pricing |
| Pro $100 ($100/mo) | No 5-hour limit; included usage allowance still applies. Check Settings > Usage for remaining allowance and reset times | Exact allowance not published by OpenAI | No universal subscription token cap published | Check reset time and options offered to your account | OpenAI Work and Codex pricing |
| Pro $200 ($200/mo) | No 5-hour limit. Current allowance unquantified; eligible accounts keep previous allowance through October 29, 2026. Check Settings > Usage for remaining allowance and reset times | Exact allowance not published by OpenAI | No universal subscription token cap published | Check reset time and options offered to your account | OpenAI Pro tiers and transition |
| Pro $500 ($500/mo) | No 5-hour limit. Highest included Pro usage; numeric estimates not published. Astra Ultrafast included. Check Settings > Usage for remaining allowance and reset times | Exact allowance not published by OpenAI | No universal subscription token cap published | Check reset time and options offered to your account | OpenAI Pro tiers and transition |
| Business Standard ($20–$25/seat/mo) | 5–3,000 (Codex Usage summary of OpenAI's model-specific estimates) local messages / 5h | Exact allowance not published by OpenAI | No universal subscription token cap published | Check workspace allowance, credits and spend controls | OpenAI Business billing and seats |
| Business Premium ($100–$125/seat/mo) | No 5-hour limit (weekly reset schedule) | Exact allowance not published by OpenAI | No universal subscription token cap published | Check workspace allowance, credits and spend controls | OpenAI Business billing and seats |
| Enterprise (Custom contract) | Contract-specific; flexible pricing has no fixed rate limits | Contract-specific | No universal subscription token cap published | Check workspace allowance, credits and spend controls | OpenAI Work and Codex pricing |
Reset timing and enterprise seat management are covered in the full Codex usage limits table.
OpenAI is rolling out GPT-6 Sol and GPT-6 Luna. The table includes their published estimates. For the newer Sol model, see GPT-6.1 Sol availability and usage limits; OpenAI now publishes a local-message estimate for Plus and Business Standard. Availability depends on your account, client and workspace.
Wondering what each subscription costs or evaluating whether Plus, Pro, or Business makes economic sense? Check our comprehensive OpenAI Codex pricing and plan costs guide.
When Your Usage Resets
When Does Your Codex Usage Reset?
Where both limits apply, two independent clocks run at once: a 5-hour window that caps bursts, and a separate weekly window that caps total volume over the weekly usage period. Which one you reach determines how long you wait.
The 5-Hour Window
Under OpenAI's documented rule (Help Center #20001516), a new five-hour window starts when you send your first message after the previous window ends. It is not continuous per-token replenishment: the window runs for 5 hours from that first prompt, after which your short-term limit clears.
Weekly Plan Pool
A separate weekly window governs cumulative volume across 7 days. Even if your 5-hour allowance resets, actions draw from your shared weekly pool. Running /status reports the exact reset time and percentage remaining for both clocks.
Reset mechanics verified 2026-09-27. Business Premium has no five-hour limit.
Plus uses five-hour and weekly windows. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex, but still have included usage allowances. Check Settings > Usage for remaining allowance and reset times. For live reset timestamps, run /status in the CLI or check your ChatGPT usage settings. For full reset mechanics and historical schedule shifts, read our deep dive on when your Codex usage resets.
First-Party Research: Why Token Counts Don't Equal Quota Drain
Developers often assume that local token logs (such as those recorded in ~/.codex) map directly to subscription quota drain. In our exploratory controlled measurements, observed quota meter movement varied across model tiers and execution modes rather than tracking client token totals linearly:
Across two controlled benchmark tasks, the displayed 5-hour meter moved 0pp for Luna, 2pp and 3pp for Sol, and 4pp and 6pp for Astra. On Task 1 (112 tests), Astra evaluated the fewest client-reported input tokens (113k vs Sol's 172k and Luna's 225k) yet moved the meter the most. These single comparisons illustrate that client token counts do not predict quota drain, but they do not establish a universal model multiplier or fixed conversion ratio.
In one controlled GPT-5.6 Sol Medium task, Fast moved the displayed 5-hour usage meter by 8 percentage points versus 2 points for Standard while completing the task about 10% faster. This single comparison does not establish a universal Fast-mode usage multiplier.
What Happens When You Hit the Limit
What Happens When You Hit the Limit?
When you hit your limit, your session is temporarily paused. OpenAI does not delete your pending task, discard uncommitted code, or wipe conversation history.
These two panels are schematic examples of the two states, not copies of what Codex prints. In the first, new prompts stay paused until the reset window clears; run /status to see when that is.
A usage limit does not itself modify your local files, Git branch, or staged changes.
If you reach a limit during an active turn, Codex can continue working on that turn, subject to fair-use limits. Afterward, wait for the displayed reset or use any available credit, reset, or upgrade option.
If urgent, you can switch your CLI profile to metered API keys to bypass plan caps.
Both cases, and how to get moving again, are covered in why am I rate limited and what happens now.
Selected personal Plus and Pro accounts may also be offered a Luna Reserve fallback, a separate GPT-5.6 Luna allowance that becomes relevant after regular usage is exhausted.
Codex Pulse
Observed
199 GitHub community reports observed by Codex Pulse.
0 change in reports since the previous scheduled observation.
Usage-related issues in openai/codex, within Pulse's 14-day creation window. Counts describe this source set, not affected users or platform-wide OpenAI behavior.
See full Codex Pulse evidence →How to Keep Usage Visible While You Work
Stop guessing your quota mid-refactor.
Codex Usage is a free local tool. It reuses your already-authenticated Codex CLI to track every limit reported by your account, check your most pressured window, and monitor reset times via a native macOS menu bar companion, terminal CLI, or local browser dashboard.
- No separate account, OAuth, or API key
- ~30-second install, no sudo
- Local browser dashboard and menu bar companion refresh every 60s
- Runs locally; session tracking is in memory only, no telemetry sent back to us