Codex

ChatGPT "Codex and Work mode usage limit reached"

Last checked

The error

Codex and Work mode usage limit reached — upgrade your plan to continue now, or wait until the limit resets on Oct 10, 15:45

The reset time at the end is yours. Seen in the ChatGPT app in Work mode and in Codex, because both draw from the same allowance.

In ChatGPT, Work mode (the Chat or Work toggle above the composer) and Codex share one usage allowance. Plus and Business have an approximately five-hour window plus weekly limits; Pro plans currently have no five-hour limit. When you use it up, both Work and Codex stop until the reset time in the message, while normal Chat keeps working. Check the exact reset on the Codex usage page or with /status in the Codex CLI, switch to Chat for anything that is not agent work, or buy credits on Plus and Pro.

Why it happens

OpenAI meters ChatGPT Work and Codex together. Per OpenAI's Codex pricing docs, Work inside ChatGPT uses the same pricing, credits and usage limits as Codex, so a long Codex session can leave no Work allowance, and the other way round.

  1. The five-hour window ran out. On Plus and Standard Business, OpenAI publishes estimates of local messages per five-hour window by model, from 5 to 45 for GPT-6 Astra up to 350 to 3,000 for GPT-6 Luna, so the model you pick changes how fast it drains.
  2. A weekly limit. OpenAI says weekly limits may also apply, without publishing fixed numbers.
  3. Heavy Codex use in the same window: long sessions, parallel subagents, or a large model at high effort. Subagents inherit the parent's model and effort unless agents.default_subagent_model is set.
  4. December changes. OpenAI's help centre says Plus, Pro and Business will share one allowance across ChatGPT and Codex from December, measured in tokens, with Instant chat excluded.

The fix

  1. 1 Read the reset time at the end of the message, or open the Codex usage page (chatgpt.com/codex/settings/usage). In the Codex CLI, run /status.
  2. 2 Switch the composer toggle to Chat for anything that does not need Work. Chat has its own limits.
  3. 3 If a task was mid-turn when you hit the limit, let it finish: OpenAI's docs say the agent can continue working on that turn.
  4. 4 On Plus or Pro, buy credits to keep going before the reset.
  5. 5 To make the next window last, set a lighter subagent model (agents.default_subagent_model, for example gpt-6-luna for simple work) and start large models at low effort.
  6. 6 For automation that runs all day, use an API key with a spend cap instead of a ChatGPT plan.
/status

Still failing?

  • The limit resets but you hit it again within minutes. Something is running in the background: check for parallel Codex tasks or cloud agents.
  • You are on Pro and see a five-hour message. Check which workspace you are signed into, since a Business workspace has its own limits.

Related errors

Full guideChatGPT and Codex Usage Limits by Plan: What Is SharedEvery error, one pageClaude Code and ChatGPT rate limit alternatives

Hit a different error?

Paste any agent error and get the cause and fix in seconds.

Open the decoder

Frequently asked questions

What is the ChatGPT 5-hour limit?

Plus and Business plans get an approximately five-hour usage window for Work mode and Codex together, plus weekly limits. When it runs out, Work and Codex pause until the reset time shown; Chat keeps working.

When does the Codex and Work mode limit reset?

At the time printed at the end of the message. The Codex usage page and /status in the Codex CLI show the current reset time too.

Does Pro raise the Work limit?

Per OpenAI's Codex pricing docs, Pro plans currently have no five-hour limit, though weekly limits can still apply. That may change when the shared allowance starts in December.

Are ChatGPT and Codex limits shared?

Work mode and Codex already share one allowance. OpenAI's help centre says that from December, Plus, Pro and Business plans will share one token-measured allowance across ChatGPT and Codex, with Instant chat excluded.

Can I keep working while I wait?

Yes. Use Chat for non-agent work, buy credits on Plus or Pro, or run the job through the API with your own key.

Stop firefighting agent errors

Decoding errors one at a time is the manual version of what BetterClaw automates. Run your agents on a no-code AI agent platform with managed models, retries and config validation built in.

Deploy Your Agent — Start FreeBook a Demo

Free plan available · Pro $49/mo · BYOK · 7-day free trial