Codex Plans and Usage Limits, Compared
Codex on every ChatGPT plan, and on an API key
| Free — where Codex runs | The ChatGPT desktop app, subject to rollout. |
|---|---|
| Free — models | GPT-6 Luna at Standard speed. |
| Free — usage | Limited Codex access; ChatGPT's comparison table marks it "Limited". |
| Go — where Codex runs | The ChatGPT desktop app, subject to rollout. |
| Go — models | GPT-6 Luna at Standard speed. |
| Go — usage | Limited Codex access, the same Codex line as Free on the Codex pricing page. |
| Plus — where Codex runs | Codex on the web, in the CLI, in the IDE extension and on iOS, plus the desktop app. |
| Plus — models | GPT-6.1 Sol (rolling out), GPT-6 Sol and GPT-6 Luna, with GPT-6 Astra listed on ChatGPT's plan table; GPT-5.6 models remain during the rollout. |
| Plus — usage | Five-hour usage windows, with weekly limits that may also apply. Purchased ChatGPT credits extend usage past the included allowance. |
| Pro — where Codex runs | Everything in Plus. |
| Pro — models | Everything in Plus with expanded usage; GPT-6 Astra with Ultrafast on the top Pro tier only. |
| Pro — usage | Three tiers with increasing included usage. Pro plans currently have no five-hour limit; included usage is used first, then purchased credits. |
| Business — where Codex runs | ChatGPT and Codex across desktop and mobile apps, the CLI, the IDE extension and Codex cloud. |
| Business — models | GPT-6.1 Sol (rolling out), GPT-6 Sol and GPT-6 Luna; Astra Ultrafast is not available. |
| Business — usage | Standard Business seats share Plus's five-hour estimates. Usage-based Codex-only seats need workspace credits from the first task. |
| Enterprise & Edu — where Codex runs | Everything in Business. |
| Enterprise & Edu — models | As Business, plus Astra Ultrafast on eligible plans. GPT-6.1 Sol stays off until an administrator enables it. |
| Enterprise & Edu — usage | With flexible (credit-based) pricing there are no fixed rate limits — usage scales with credits. Without it, per-seat limits match Plus for most features. |
| API key — where Codex runs | The CLI, the SDK and the IDE extension — no cloud features (no GitHub code review, Slack or Codex cloud). |
| API key — models | Whatever API models your key can call. |
| API key — usage | Billed at API token prices. ChatGPT credit multipliers do not apply, and ChatGPT sign-in retirements (GPT-5.5, GPT-5.4) do not affect it. |
The Pro tiers, by rank
| Pro, entry tier | The smallest included usage of the three. No Ultrafast. |
|---|---|
| Pro, middle tier | More included usage than the entry tier. Reopened to new subscribers with a lower allowance than it previously carried. No Ultrafast. |
| Pro, top tier | The highest included usage of the three, and the only Pro tier with Astra Ultrafast. |
| Grandfathered middle-tier allowance | Kept through 2026-10-29 by subscribers active at the eligibility cutoff, then the lower allowance at the same price |
OpenAI's estimate of local messages per 5-hour window — ranges, not limits
| GPT-6 Astra | Plus 5–45 · Standard Business 5–45 local messages per 5 hours |
|---|---|
| GPT-6.1 Sol | Plus 15–160 · Standard Business 15–160 local messages per 5 hours |
| GPT-6 Sol | Plus 15–150 · Standard Business 15–150 local messages per 5 hours |
| GPT-6 Luna | Plus 350–3,000 · Standard Business 350–3,000 local messages per 5 hours |
Speed modes inside a ChatGPT plan
| Fast | 2.5× Standard against included usage · 2× against purchased credits |
|---|---|
| GPT-6 Astra Ultrafast | 8× Standard against included usage · 6× against purchased credits |
Feature availability by plan (Free and Go are not columns in OpenAI's matrix)
| Codex CLI | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
|---|---|
| IDE extension | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
| Codex SDK, codex exec and scripted workflows | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
| Codex cloud | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key — |
| Codex access tokens for trusted automation | Plus — · Pro — · Business Yes · Ent/Edu Yes · API key — |
| GPT-6 Sol and Luna | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
| Fast mode | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
| Astra Ultrafast | Plus — · Pro Yes · Business — · Ent/Edu Yes · API key Yes |
| Web search | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
| Local code review with /review | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
| Auto-review for approval requests | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Yes |
| GitHub code review and automatic PR reviews | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key — |
| Slack and Linear cloud integrations | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key — |
| Plugins | Plus Yes · Pro Yes · Business Yes · Ent/Edu Yes · API key Limited |
| Computer use | Plus Limited · Pro Limited · Business Limited · Ent/Edu Limited · API key Limited |
| SAML SSO, MFA and workspace user management | Plus — · Pro — · Business Yes · Ent/Edu Yes · API key — |
| No training on API or business data by default | Plus — · Pro — · Business Yes · Ent/Edu Yes · API key Yes |
| Compliance API and audit logs | Plus — · Pro — · Business — · Ent/Edu Yes · API key — |
Which ChatGPT plan you're on decides more about Codex than most people expect: where it runs, which models appear in the picker, how usage is metered, and whether the fastest tier is available at all. The tables above cover every plan except on price — this site prints no plan prices, for reasons the subscription policy page explains — and the notes below are what tends to surprise people.
Free and Go are desktop-app plans
OpenAI's Codex pricing page lists only the ChatGPT desktop app for Free and Go, with GPT-6 Luna at Standard speed and usage marked "Limited". It doesn't confirm CLI access for either, and it publishes no Codex allowance figure, so neither does this site.
Usage windows, and what the estimates mean
Plus and standard Business seats meter usage in 5-hour windows, with weekly limits that may also apply; usage-based Codex-only Business seats run on workspace credits from the first task. OpenAI publishes estimates of local messages per window for each model, as ranges, and says explicitly that they aren't fixed message limits — the spread reflects how much a task's size changes what one message costs. Cloud tasks may use more of the allowance than local ones. Pro currently has no 5-hour limit, and OpenAI publishes no comparable estimate for it.
The Pro ladder
Pro is now a set of tiers ranked by included usage. Only the top tier includes GPT-6 Astra Ultrafast, and buying credits on a lower tier doesn't unlock it. The middle tier reopened to new subscribers with a smaller allowance than it used to carry; subscribers who qualify keep the previous allowance through October 29, 2026, then move to the lower one at the same price.
Speed costs more against the plan than against credits
Fast and Ultrafast both consume included usage at a steeper multiple than they consume purchased credits. The speed table above shows both, and why Fast mode burns your plan faster explains what that means in practice.
When the allowance runs out
On plans that support it, purchased credits extend usage past the included allowance — included usage is spent first, then credits. The Codex credits calculator prices a task in credits, and the subscription vs API cost calculator compares a plan with paying per token.
Verified 2026-10-01 against https://learn.chatgpt.com/docs/pricing.
Could not confirm: Plan prices (by policy — see SUBSCRIPTION_INPUT_POLICY), and whether Free and Go include the Codex CLI: the Codex pricing page names only the desktop app for those two plans.
Checked: https://learn.chatgpt.com/docs/pricing · https://chatgpt.com/pricing/ · https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers · https://learn.chatgpt.com/docs/models · https://help.openai.com/en/articles/11481834-chatgpt-rate-card-business-enterpriseedu-credit-based-pricing
Verified 2026-10-01 against https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers.
Could not confirm: How much included usage each Pro tier carries in absolute terms: OpenAI describes only their order.
Checked: https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers · https://learn.chatgpt.com/docs/pricing · https://learn.chatgpt.com/docs/agent-configuration/speed
Verified 2026-10-01 against https://learn.chatgpt.com/docs/pricing.
Could not confirm: Estimates for Pro (which currently has no five-hour limit), for cloud tasks (which "may use more of your allowance"), and for the weekly limits that "may also apply" — none is published as a number.
Checked: https://learn.chatgpt.com/docs/pricing
Verified 2026-10-01 against https://learn.chatgpt.com/docs/agent-configuration/speed.
Checked: https://learn.chatgpt.com/docs/agent-configuration/speed · https://learn.chatgpt.com/docs/pricing
Verified 2026-10-01 against https://learn.chatgpt.com/docs/pricing.
Could not confirm: "Limited" cells carry regional or plugin limits that OpenAI describes in footnotes, not per plan.
Checked: https://learn.chatgpt.com/docs/pricing