Codex Usage Limits: Check Remaining Usage & Reset Times

Check Codex usage limits and reset times on the usage dashboard or via /status, compare Plus, Pro & Business allowances, and keep working with an API key.

UniRoute · October 22, 2026

If Codex paused your coding session with a usage limit notice, here is where to check your remaining allowance and reset time, how much usage each ChatGPT plan includes, and the three ways to keep working right now.

1. Where to Check Remaining Usage & Reset Time

Web Usage Dashboard: Open the official Codex usage dashboard (chatgpt.com/codex/settings/usage) to view your current allowance, credit balance, and reset window (Codex Pricing, verified 2026-10-27).
Inside Codex CLI: Run /status in an active CLI session to inspect your remaining limits (Codex Pricing, 2026-10-27), or run codex login status in your terminal to check whether the CLI is authenticated via ChatGPT sign-in or an API key (Codex Authentication, 2026-10-22).

Important: Do not run codex logout or delete ~/.codex/auth.json to troubleshoot a limit message. Usage counters are tracked server-side; deleting local credentials only forces you to sign in again.

2. Codex Plan Allowances & Speed Mode Multipliers (Official Estimates, 2026-10-27)

Per OpenAI's official Codex Pricing documentation (verified 2026-10-27), ChatGPT plans are priced at Free ($0/mo), Go ($8/mo), Plus ($20/mo), Pro (from $100/mo: $100, $200, or $500/mo tiers), and Business ($20/user/mo billed annually for 2+ users, or $25/user/mo billed monthly), alongside Enterprise and Edu plans.

Local Messages per 5-Hour Period (Plus & Standard Business Estimates)

Model	Local Messages per 5-Hour Period (Estimate)
GPT-6 Astra	5–45
GPT-6.1 Sol	15–160
GPT-6.1 Terra	35–285
GPT-6 Sol	15–150
GPT-6 Luna	350–3,000

Note (2026-10-27 official pricing page): The numbers above are estimates, not fixed limits—actual messages depend on prompt complexity, repository context, and tool calls. Pro currently has no 5-hour limit; weekly limits and fair-use policies may also apply across plans.

Speed Mode Consumption Multipliers (Relative to Standard Mode)

Toggling /fast in the CLI changes inference speed preference (Speed configuration, 2026-10-22) and consumes allowance or credits at a higher multiplier (Codex Pricing, 2026-10-27):

Speed Mode	Included Subscription Usage	Purchased Credits & Enterprise Pay-as-You-Go
Fast	2.5x	2x
Ultrafast (GPT-6 Astra / GPT-6.1 Sol)	8x	6x

3. Three Options When You Hit Codex Usage Limits

According to Codex Pricing (2026-10-27), when your account reaches its usage window limit, you have three options:

Wait for your reset window (active turns can finish): An in-progress turn can continue to completion (subject to fair use). Turn off /fast and split large multi-file tasks into smaller prompts once your window resets.
Purchase extra ChatGPT credits: Plus and Pro users (as well as flexible Business, Edu, and Enterprise workspaces) can purchase additional credits to continue working before the next reset.
Switch to an API key for local chats: As OpenAI's official pricing page states:

"All users may also run extra local chats using an API key, with usage charged at standard API rates."

How to Keep Working Immediately with an API Key (OpenAI Platform or UniRoute)

You can connect local Codex CLI to a direct OpenAI Platform API key or route local sessions through UniRoute (create a UniRoute API key) on pay-as-you-go billing via ~/.codex/config.toml (Alternative model providers):

npx --yes --package=https://uniroute.app/community/uniroute-setup-0.1.2.tgz uniroute-setup models
npx --yes --package=https://uniroute.app/community/uniroute-setup-0.1.2.tgz uniroute-setup configure codex --model MODEL_ID

UniRoute GPT Pricing (as of 2026-10-27): As of 2026-10-27, on UniRoute's discounted channel GPT models are as low as ~28% of official API rates (e.g., GPT 6 Sol is $0.56 / $2.80 per 1M input/output tokens vs. $2 / $10 official, and GPT 5.6 Terra is $0.56 / $3.36 vs. $2 / $12 official); when the discounted channel is unavailable, requests fail over automatically to the standard channel at official-level prices (how channels work; see UniRoute Pricing).
How to estimate per-request cost: API billing follows Cost = (Input Tokens / 1,000,000 × Input Rate) + (Output Tokens / 1,000,000 × Output Rate). (Hypothetical example: a single request with 20,000 uncached input tokens and 2,000 output tokens on GPT 6 Sol at $0.56 / $2.80 per 1M tokens costs (20,000 / 1,000,000 × $0.56) + (2,000 / 1,000,000 × $2.80) = $0.0168.)
Official API-Key Boundaries (learn.chatgpt.com/docs/pricing, 2026-10-27): API keys work only in local environments (Codex CLI, SDK, or IDE extension); cloud-based features (such as GitHub code review or Slack integrations) require ChatGPT sign-in and are not available with an API key, and available models depend on the API models supported by your key. See our Codex CLI setup documentation, Connect Codex to UniRoute tutorial, or How to Install Codex CLI.

Identify Which Restriction Paused Your Session

Not every interruption comes from your personal 5-hour or weekly plan window. Match what you see in your client against the three categories below:

Restriction Category	Typical Signal	How to Verify	Practical Next Step
1. Personal Plan Window	Client blocks new turns on your signed-in personal account	Check the usage dashboard or `/status`	Wait for reset, buy credits, or switch to an API key above
2. Workspace Permissions & Spend Controls	Prompts are blocked on a team account even though your seat is active	Ask a workspace owner or admin to review workspace settings	Admin updates role access, shared credit pool, or per-user spend cap
3. API or Provider Request Error	Request fails when running with an API key or custom model provider	Inspect the exact response error code and message, active auth method, configured endpoint, and provider dashboard/docs	Address the confirmed cause (such as an invalid key, endpoint mismatch, rate policy, or depleted balance)

Team Workspaces vs. Personal Plan Limits

On Business, Enterprise, and Edu plans, seat access, role permissions, and workspace spend controls (2026-10-22) are separate checks; shared or per-user credit caps can pause Codex without governing standalone API billing. Upgrading a personal account will not lift an organization cap—ask a workspace admin to inspect team settings.

Workflow Adjustments & Billing Separation

1. Monitor Active Work and Narrow Task Scope

Do not assume an in-progress turn will follow a fixed blocking sequence or that all accounts reset on the same clock hour. Monitor the active turn in your client, because fair-use restrictions, tool failures, network drops, or provider errors can still interrupt a session. Before retrying a large task, break multi-file refactors into smaller, verifiable steps, start a fresh session when switching to an unrelated problem, and leave /fast off unless low latency is essential.

2. Independent Billing Channels

Keep three operational boundaries clear when switching between ChatGPT sign-in and API keys:

Separate accounts and balances: Your ChatGPT subscription, your OpenAI Platform API project, and your gateway balance are billed independently. Connecting an API key never resets or extends your ChatGPT subscription window.
Local vs. cloud availability: API keys cover local workflows (CLI, SDK, IDE extension); cloud features require ChatGPT sign-in (feature availability, 2026-10-27).
Configuration boundary: OpenAI documents the config.toml schema for custom responses-compatible model providers, but does not certify third-party gateway supply, quotas, or model availability.

Frequently Asked Questions

Why shouldn't I delete `~/.codex/auth.json` when troubleshooting a limit?

~/.codex/auth.json only stores your local login credentials. Quota counters are tracked on the server side by your account or API provider, so deleting the file simply forces you to sign in again without changing your usage state.

Why did my request fail right after I switched to an API key or custom provider?

Inspect the exact error code and message returned in the terminal before choosing a fix. Possible causes include an unexported environment variable, a mismatched endpoint URL, or a provider-side project restriction.

Does routing local CLI sessions through an API gateway change my existing account billing?

Model requests routed to a configured API provider follow that provider's account and pricing terms, while your ChatGPT subscription allowance remains separate.