ChatGPT Codex limits in 2026 are estimate ranges, not one fixed message cap. OpenAI's current table covers the GPT-5.6 family: Sol for deeper reasoning, Terra as the everyday default, and Luna for faster, higher-volume work. Check Codex Settings > Usage, or run /status in an active Codex CLI session, for your actual remaining allowance.
The practical version: Plus supports a few focused coding sessions each week, Pro $100 gives 5x the Plus usage, and Pro $200 gives 20x. Business and Enterprise are mainly workspace decisions. If limits repeatedly interrupt real work, wait for the reset, buy credits if eligible, or move up one tier.
Quick answer
- Plus GPT-5.6 ranges are 10-100 for Sol, 25-200 for Terra, and 250-2,000 for Luna per five hours.
- Pro $100 is the 5x tier. Pro $200 is the 20x tier.
- Local messages and cloud chats share your allowance; additional weekly limits may apply.
- ChatGPT Work and Codex share the same agentic usage and credit pool.
Official GPT-5.6 Codex limit estimates
OpenAI's Codex pricing page lists these local-message planning ranges per five-hour window for GPT-5.6. Checked September 9, 2026. Other models have separate rows in the official table; use your usage dashboard for your available models and actual limits.
| Model | Plus / Standard Business | Pro 5x | Pro 20x |
|---|---|---|---|
| GPT-5.6 Sol | 10-100 / 5h | 50-500 / 5h | 200-2,000 / 5h |
| GPT-5.6 Terra | 25-200 / 5h | 125-1,000 / 5h | 500-4,000 / 5h |
| GPT-5.6 Luna | 250-2,000 / 5h | 1,250-10,000 / 5h | 5,000-40,000 / 5h |
Standard Business has the same published ranges as Plus; Business $100 uses the Pro 5x estimates. Enterprise and Edu without flexible pricing have the same per-seat limits as Plus for most features. Enterprise and Edu workspaces with flexible pricing have no fixed rate limits; usage scales with credits.
GPT-5.5 and GPT-5.4 limits
OpenAI's pricing page still lists these older-model ranges. However, its rate card states that GPT-5.4 and GPT-5.4 mini retired from Codex for users signed in with ChatGPT on August 31, 2026. Those two rows are reference-only for that sign-in mode, not a promise of availability. OpenAI API access and Codex use with your own API key are unaffected; OpenAI recommends GPT-5.6 Terra and Luna respectively for ChatGPT-signed-in users.
| Model | Plus / Standard Business | Pro 5x | Pro 20x |
|---|---|---|---|
| GPT-5.5 | 15-80 / 5h | 75-400 / 5h | 300-1,600 / 5h |
| GPT-5.4 | 20-100 / 5h | 100-500 / 5h | 400-2,000 / 5h |
| GPT-5.4 mini | 60-350 / 5h | 300-1,750 / 5h | 1,200-7,000 / 5h |
What about cloud tasks and code reviews?
OpenAI says cloud chats on ChatGPT plans use GPT-5.6 Sol and may consume more allowance than local messages. Local messages and cloud chats share your plan's allowance. Do not use older numeric cloud-task or code-review ranges for current planning.
Plus includes automatic code review and other cloud integrations. The rate card says GitHub code review uses GPT-5.3-Codex. Reviews run locally or outside GitHub count toward general usage limits. Availability and credit use depend on your plan, workspace, feature and model, not just the local-message table.
Treat every range as a planning estimate. Model choice, large repositories, long prompts, tool use, retrieval, caching, and long outputs can change how quickly you use the allowance.
Which ChatGPT plan should you choose for Codex?
| Plan | Choose it when |
|---|---|
| Free / Go | You only want to test Codex. |
| Plus | You use Codex for small fixes, scripts, and occasional coding help. |
| Pro $100 | Plus limits interrupt real work a few times a week. |
| Pro $200 | You use Codex heavily every day or across multiple projects. |
| Business / Enterprise | You need team billing, admin controls, SSO, compliance, or workspace credits. |
| API key | You need automation, CI, or scripted local jobs. |
For broader plan differences, see our ChatGPT Plus vs Pro vs Business guide.
What counts against Codex limits?
Codex counts toward agentic usage. The exact drain depends on the work, not just the number of prompts.
- Model: larger models usually consume more allowance or credits.
- Context: large repos, long chats, and big AGENTS.md files cost more.
- Surface: Codex, ChatGPT Work, and supported agentic features can draw from the same usage and credit pool.
- Output: long edits, long explanations, and multi-file changes use more.
What happens when you hit the Codex limit?
| Plan | Next step |
|---|---|
| Free / Go | Wait for the displayed reset, upgrade, or buy credits if Add credits is available to your account. |
| Plus / Pro | Buy credits if eligible, wait for reset, or upgrade. |
| Business / Edu / Enterprise | Use workspace credits and admin controls. |
| API key | Usage is billed separately at API rates. |
How do Codex credits work in 2026?
ChatGPT Work and Codex credit rates are token-based. Eligible personal-plan users spend included usage first, then purchased credits. Standard Business seats also use included limits first; usage-based Codex-only seats require workspace credits from first use. Enterprise and Edu flexible-pricing workspaces scale with credits. A small subset of Enterprise customers may still use the legacy rate card.
Bigger context, more tools, longer output, and fast mode usually mean higher credit usage. These are the published standard-speed rates per million tokens, checked September 9, 2026. GPT-5.6 Sol's promotional pricing for eligible purchased-credit usage runs at least through November 21, 2026; it does not change included five-hour or weekly limits. Older-model rates below do not override the sign-in availability notes above.
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-5.6 Sol | 100 credits / 1M | 10 credits / 1M | 500 credits / 1M |
| GPT-5.6 Terra | 50 credits / 1M | 5 credits / 1M | 300 credits / 1M |
| GPT-5.6 Luna | 5 credits / 1M | 0.5 credits / 1M | 30 credits / 1M |
| GPT-5.5 | 125 credits / 1M | 12.50 credits / 1M | 750 credits / 1M |
| GPT-5.4 | 62.50 credits / 1M | 6.25 credits / 1M | 375 credits / 1M |
| GPT-5.4 mini | 18.75 credits / 1M | 1.875 credits / 1M | 113 credits / 1M |
OpenAI's rate card estimates roughly 5-30 credits for a typical GPT-5.6 Sol Codex task, but actual use depends on the input, cached input, output, tools, and mode used.
How to make Codex limits last longer
- Narrow the task: ask for one clear change at a time.
- Trim context: remove old logs, repeated instructions, and irrelevant files.
- Keep AGENTS.md focused: long instructions are injected often.
- Disable unused MCP servers: extra tools add overhead.
- Use smaller models when suitable: only if they are available in your Codex picker.
- Watch usage: check Codex Settings > Usage or /status in the CLI.
Bottom line
Use Plus for a few focused Codex sessions each week. Move to Pro $100 if Plus repeatedly interrupts real work, and use Pro $200 for heavy daily usage. Choose Business or Enterprise for team controls, and use API billing for automation. Remember that ChatGPT Work and Codex share usage.
Frequently Asked Questions
Does ChatGPT Plus include Codex?
Yes. OpenAI says Codex is included across Free, Go, Plus, Pro, Business, Edu, and Enterprise plans. Plus is designed for a few focused coding sessions each week.
What is the ChatGPT Plus Codex limit?
OpenAI gives estimate ranges, not one fixed cap. For Plus, GPT-5.6 Sol is estimated at 10-100 local messages per five hours, Terra at 25-200, and Luna at 250-2,000. Local messages and cloud chats share your allowance; additional weekly limits may apply.
What is the difference between Pro $100 and Pro $200 for Codex?
Pro $100 is the 5x tier, with GPT-5.6 ranges from 50-500 for Sol to 1,250-10,000 for Luna per five hours. Pro $200 is the 20x tier, ranging from 200-2,000 for Sol to 5,000-40,000 for Luna.
Can I buy more Codex usage after I hit the limit?
Eligible Plus and Pro users can buy ChatGPT credits. Credit purchases are also available to a limited group of Free and Go users when Add credits appears in their account. Business, Edu, and Enterprise workspaces may use workspace credits, depending on their pricing setup.
Are GPT-5.6 cloud tasks and code reviews available?
OpenAI says cloud chats on ChatGPT plans use GPT-5.6 Sol and share the allowance with local messages. Plus includes cloud integrations such as automatic code review. GitHub code review uses GPT-5.3-Codex; reviews run locally or outside GitHub count toward general usage limits. Availability still depends on your plan and workspace.
Do API key Codex tasks use my ChatGPT Codex limit?
No. API key usage is billed separately at API rates and is better for CI, automation, and scripted workflows.
Where can I see my remaining Codex usage?
Use Codex Settings > Usage. In an active Codex CLI session, OpenAI also says you can use /status.
What if the docs mention a model that is not in my Codex picker?
Use your Codex picker and usage dashboard as the practical source of truth. Availability can vary by plan, workspace, region, or rollout.
Sources
Sources checked on September 9, 2026. The five-hour ranges and token rates come from OpenAI's Codex pricing page, cross-checked against its rate card. Personal-plan credit eligibility and workspace billing rules differ; check the source for your plan and your usage dashboard before purchasing.
- OpenAI Codex pricing and usage limits: https://learn.chatgpt.com/docs/pricing
- Using Codex with your ChatGPT plan: https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan
- About ChatGPT Pro plans: https://help.openai.com/en/articles/9793128-about-chatgpt-pro-plans
- Using credits for flexible usage in ChatGPT: https://help.openai.com/en/articles/12642688-using-credits-for-flexible-usage-in-chatgpt-freegopluspro
- ChatGPT rate card (Business, Enterprise/Edu credit-based pricing): https://help.openai.com/en/articles/11481834