What Happened
OpenAI restored the 5‑hour daily limit on Codex and Work calls for ChatGPT Plus users. The limit had been lifted temporarily to let developers experiment with new AI‑agent features. Now, each Plus subscriber can make only five hours of combined Codex or Work calls per day before the service throttles or blocks further requests until the reset.
The change was announced in a brief blog post on OpenAI’s site and followed up by a 9to5Mac article that warned developers building large‑scale AI workflows about the impact. No new pricing tiers appeared; the quota simply returned to its former state.
Why This Matters for Builders
- Predictable Compute Budgets: The 5‑hour cap forces teams to budget Codex or Work usage each day, encouraging tighter prompts or batching.
- Cost Management: A hard cap prevents runaway AI calls that could trigger unexpected charges. Builders can set clear cost ceilings in CI/CD pipelines.
- Workflow Design: Scripts that rely on continuous Codex calls—like real‑time code generation or dynamic task planning—must fit within the quota. Adding caching or pre‑computing snippets can help.
- Reliability and Scaling: High‑availability AI agent deployments face a new bottleneck. Architects may need fallback logic or queueing to handle throttling gracefully.
FAQ
Q: Can I request a higher limit for my production environment?
A: OpenAI does not offer custom limits for Plus users. Enterprise plans or dedicated API access may provide higher quotas. Contact sales for details.
Q: Will the limit apply to all API calls, or only to Codex and Work endpoints?
A: The limit targets only Codex and Work usage. Other endpoints, such as the standard ChatGPT API, remain unaffected.
Q: How can I monitor when I’m approaching the 5‑hour threshold?
A: Use the OpenAI dashboard or the usage endpoint to track daily consumption. Integrate alerts into your monitoring stack to trigger at 80% of the quota.
Originally published on Automations Cookbook.
