Pro $200 is available for new subscriptions with a lower allowance for non-grandfathered accounts; OpenAI has not published its replacement multiplier. Eligible grandfathered subscribers keep the previous allowance through October 29, 2026 while subscribed, then move to the lower allowance. Earlier published Pro 5x and Pro 20x message estimates are historical evidence, not current five-hour caps or proof of current Pro $200 entitlement. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex, but still have included usage allowances. Pro $500 has the highest included Pro usage and Astra Ultrafast, with no published multiplier or model-specific message estimate. OpenAI Pro tiers.
Yes. GPT-6 Astra is usage-limited in Codex, and how much usage you get depends on your ChatGPT plan. Astra draws from your existing Work and Codex allowance rather than having an isolated message quota. OpenAI publishes estimated 5-hour local-message ranges for Plus and Business Standard. Pro $100, $200, $500 and Business Premium have no five-hour usage limit; the Pro numbers below are earlier published references, not current caps:
| Plan Tier | 5-Hour Message Estimate | Astra Pool Access | Weekly Quota |
|---|---|---|---|
| ChatGPT Plus ($20/mo) | 5 to 45 local messages (official estimate) | Limited Astra usage in Work and Codex | Not published by OpenAI |
| ChatGPT Pro $100 ($100/mo) | 25 to 225 local messages (earlier 5x estimate; no current five-hour cap) | Full existing Work and Codex allowance | Not published by OpenAI |
| ChatGPT Pro $200 ($200/mo) | 100 to 900 local messages (legacy Pro 20x estimate; no current five-hour cap) | Full existing Work and Codex allowance | Not published by OpenAI |
| Business Standard ($20–$25/mo) | 5 to 45 local messages (shares Plus estimates) | Limited Astra usage in Work and Codex | Not published by OpenAI |
| Business Premium ($100–$125/mo) | No 5-hour limit (5x Standard included usage) | Full existing Work and Codex allowance | Weekly reset schedule (exact cap not published) |
These figures are OpenAI's estimated local-message ranges, not guaranteed fixed prompt caps. Actual usage varies with factors OpenAI documents: task complexity, model behavior, reasoning effort, context size, input and output tokens, tools, and execution mode (such as Fast mode). Business Premium is structurally different from the 5-hour estimate rows: OpenAI documents no 5-hour usage limit for Premium seats, metering usage strictly under a weekly reset schedule with 5x Standard included usage.
Draws from your plan's included allowance; no standalone message quota.
Pro $100/$200 & Premium use full allowance; Plus & Standard include limited usage.
Earlier Pro references (Pro $100: 25–225; Previous Pro 20x: 100–900); Pro $100, $200, $500 and Premium have no five-hour usage limit.
Codex CLI version 0.153.0+ required for Astra model configurations.
Astra Limits Across Subscription Plans & API Billing
Developers frequently confuse metered API rate limits (such as Tier 1–5 RPM and TPM) with ChatGPT subscription limits. This table compares how Astra usage is treated under Plus, Pro $100, Pro $200, and direct API access. (For ChatGPT Business workspaces, Standard aligns with Plus limits while Premium removes the 5-hour limit and provides 5x Standard usage). GPT-6 Astra's September 3 integration is recorded in our complete Codex limit change history.
| Dimension | ChatGPT Plus ($20/mo) | ChatGPT Pro $100 ($100/mo) | ChatGPT Pro $200 ($200/mo) | OpenAI Platform API | Source |
|---|---|---|---|---|---|
| Where it draws from | Limited Astra usage drawing from existing Work and Codex allowance | Can use full existing Work and Codex allowance for Astra | Can use full existing Work and Codex allowance for Astra | Independent per-project/organization limits (RPM/TPM and monthly spend limit) | Official |
| 5-Hour Rolling Window | Est. 5–45 local messages (official estimate) | No five-hour usage limit; included allowance still applies | No five-hour usage limit; current allowance unquantified | Not applicable (rate limited by Tier RPM and TPM) | Official (estimate) |
| Weekly Window | Not published by OpenAI | Not published by OpenAI | Not published by OpenAI | Not applicable (metered monthly prepaid / org spend ceiling) | Not published |
| What happens at the limit | Active turn continues subject to fair-use limits; subsequent prompts pause | Active turn continues subject to fair-use limits; subsequent prompts pause | Active turn continues subject to fair-use limits; subsequent prompts pause | HTTP 429 Too Many Requests response | Official |
| Cost / Billing Structure | Included in $20/mo subscription (Work & Codex surfaces) | Included in $100/mo subscription (5x Plus usage) | Included in $200/mo subscription (previous 20x allowance; current multiplier unpublished) | Metered token billing ($10/M input, $50/M output standard; $20/M input, $100/M output Fast mode) | Official |
| Reset Options | 5-hour rolling recovery, weekly reset, paid instant reset (eligible personal accounts), or banked reset (if available) | Weekly reset, paid instant reset (eligible personal accounts), or banked reset (if available); no five-hour cap | Weekly reset, paid instant reset (eligible personal accounts), or banked reset (if available); no five-hour cap | Immediate upon rate limit window clear or prepaid balance top-up | Official |
Why Your Astra Allowance Disappears Faster
OpenAI says it improved Astra subscription usage for long-running, high-usage workloads when signed in with a ChatGPT account. For some users in the long tail, the same Astra work can now draw up to 3 to 4 times less subscription usage, with no change in model quality.
This does not mean every plan received 3 to 4 times more quota, and it should not be used to multiply the message estimates above. OpenAI described the improvement specifically as reduced subscription usage for some power-user workloads.
OpenAI has confirmed Astra, Sol and Luna; Sol and Luna are rolling out. Check GPT-6 Sol and Luna availability. OpenAI names five things that decide how much Astra you get:
A full test-suite generation, a multi-file refactor, or a debugging loop takes many round-trips and tool calls. A single question takes one.
Big files, wide repository context, verbose compiler logs, and long generated diffs all push up the token cost of every turn.
Higher reasoning effort means the model spends more compute planning, checking, and correcting itself before it answers.
Fast mode accelerates Astra generation faster than Standard processing (no exact multiplier published), consuming credits at a 2.5x multiplier where available (or 2x token rates on API billing). See our dedicated Codex Fast mode speed and credit guide.
Plus and Business Standard use five-hour and weekly windows; Pro $100, $200, and $500 have no five-hour usage limit but still have included allowances; Business Premium has no five-hour limit. Several long Astra sessions can stay comfortably under the 5-hour cap and still empty your weekly allowance days before it resets.
If Astra usage is falling faster than these model and workload factors explain, see how to diagnose unexpectedly fast Codex usage drain.
ChatGPT Plus vs. ChatGPT Pro for Astra
Astra access and capacity scale according to plan tier:
Limited Astra Usage
- Includes limited Astra usage in ChatGPT Work and the Codex CLI during rollout (est. 5 to 45 local messages per 5 hours).
- GPT-6 Pro in ordinary ChatGPT web chat is not included with Plus (see our breakdown of GPT-6 Pro vs GPT-6 Astra for surface access and quota distinctions).
- 5-hour and weekly allowances exhaust faster during intensive Astra tasks.
- Eligible personal accounts can purchase paid instant weekly resets or pay-as-you-go credits in settings.
Full Existing Allowance for Astra
- Pro $100 ($100/mo): Est. 25 to 225 local messages in the earlier Pro 5x reference; no current five-hour cap.
- Pro $200 ($200/mo): Est. 100 to 900 local messages in the previous Pro 20x reference; no current five-hour cap, and current multiplier unpublished.
- Both Pro $100 and Pro $200 can use their full existing Work and Codex allowance for Astra.
- Full support across Codex CLI, ChatGPT Work, and supported web surfaces with priority access during peak load.
- Qualifying accounts may receive promotional banked resets depending on specific offers and eligibility.
GPT-6 Astra vs. GPT-5.6 Sol: Which One Should Use Your Allowance?
Astra and Sol draw from the same shared Work and Codex allowance — there is no separate quota for either model. OpenAI publishes higher estimated 5-hour local-message ranges for Sol than for Astra, and Sol's metered token rates are lower, so switching a session from Sol to Astra spends your allowance faster.
| Plan Tier | GPT-5.6 Sol (5h estimate) | GPT-6 Astra (5h estimate) |
|---|---|---|
| Plus / Business Standard | 10 to 100 local messages | 5 to 45 local messages |
| Pro 5x ($100/mo) | 50 to 500 local messages | 25 to 225 local messages |
| Previous Pro 20x ($200/mo) | 200 to 2,000 local messages | 100 to 900 local messages |
| Standard Work & Codex Credits (per 1M tokens) | GPT-5.6 Sol | GPT-6 Astra |
|---|---|---|
| Input | 100 credits | 250 credits |
| Cached input | 10 credits | 25 credits |
| Output | 500 credits | 1250 credits |
These figures are OpenAI's published estimates and current credit rate card, not a fixed conversion formula for included subscription allowance. Included plan usage and metered token billing are distinct: purchased credits and eligible token-based Work and Codex activity use the rate card above, while direct API usage is billed separately in USD under API rate limits.
OpenAI describes GPT-6 Astra as its most capable model for the hardest end-to-end work, and GPT-5.6 Sol as its flagship model for complex, open-ended professional work. OpenAI also explicitly states that Astra can use Work and Codex allowance faster than Sol. The practical quota decision follows from that: use Sol where it is sufficient, and move to Astra when the additional capability is worth the higher allowance consumption. Neither model has a guaranteed cost-per-task advantage over the other — the right choice depends on whether the task needs Astra's extra depth. See the Codex recommended model guide for OpenAI's full model guidance, and the GPT-5.6 Luna usage limits guide if preserving allowance matters more than either Sol or Astra's reasoning depth.
| Experiment | Model | Correctness | Runtime | Observed 5-Hour Movement |
|---|---|---|---|---|
| Earlier paired study | GPT-5.6 Sol | 57 / 57 checks | — | 3 percentage points |
| Earlier paired study | GPT-6 Astra | 57 / 57 checks | — | 4 percentage points |
| Later frozen-task study | GPT-5.6 SolMedium reasoning | 112 / 112 checks | 92.0s | 2 percentage points |
| Later frozen-task study | GPT-6 AstraMedium reasoning | 112 / 112 checks | 109.7s | 4 percentage points |
Across two small controlled experiments, Astra produced at least as much observed 5-hour meter movement as Sol while deterministic correctness was unchanged. The size of the difference varied between experiments, so these measurements do not establish a fixed Astra-to-Sol consumption multiplier.
Control result: GPT-5.6 Luna at Medium reasoning passed 112 / 112 checks in 118.2s, with no detectable 5-hour movement at the meter's 1-percentage-point resolution.
These are Codex Usage's first-party observed measurements, not official OpenAI quota documentation. Run counts were very small, and the displayed quota meter has 1-percentage-point resolution. Do not treat these results as typical usage, statistical proof, or a conversion formula. See the full first-party measurement for the full tables and the methodology.
What OpenAI Has Not Published About Astra
OpenAI has published some of this and left the rest undocumented. Here is which is which:
What To Do When Astra Hits Your Codex Limit
If you reach a usage limit during an active turn, Codex can continue working on that turn, subject to fair-use limits. Once that turn concludes, subsequent prompts pause until allowance is restored. You have four recovery paths:
If you hit the 5-hour rolling burst cap, prompts pause until the rolling 5-hour window clears. If you exhausted your weekly ceiling, prompts remain paused until your account's scheduled weekly reset date. Check /status for live reset times.
Banked resets are promotional benefits whose eligibility and expiration depend on specific offers, plans, workspaces, and accounts. If an unused banked reset is available on your account, you can apply it in Settings > Usage to refresh your 5-hour and weekly usage windows.
Eligible customers on personal ChatGPT Plus and Pro plans can purchase an instant weekly rate limit reset in Settings > Usage or via in-app banners when hitting a limit. This pulls your normal weekly cycle forward immediately. Pay-as-you-go credits can also be added where supported.
Authenticating the Codex CLI with an OpenAI API key instructs it to route requests to the OpenAI Platform API instead of your subscription. This bypasses subscription 5-hour and weekly limits, billing your API credit balance at standard token rates ($10/M input, $50/M output standard; $20/M input, $100/M output Fast mode for gpt-6-astra).
How to Check Your Remaining Codex Quota
Because Astra draws from your overall Work and Codex allowance, you can check your live 5-hour and weekly percentages in the terminal:
You can also check the official ChatGPT usage settings page or use our free local Codex Usage desktop monitoring tool to track 5-hour and weekly percentages. Note that neither OpenAI nor the local tool attributes usage specifically to Astra on a per-prompt or token level. For every other place those numbers show up, see how to check your Codex usage.
Frequently Asked Questions
Does GPT-6 Astra have its own separate usage quota in Codex?
Astra draws from your existing Work and Codex allowance rather than having an isolated separate quota. OpenAI states that ChatGPT Plus and Business Standard include limited Astra usage, while Pro $100, Pro $200, and Business Premium can use their full existing Work and Codex allowance for Astra. On Plus and Business Standard, Astra draws from shared five-hour and weekly windows. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex, but still have included allowances. Check Settings > Usage for remaining allowance and reset times. Business Premium has no five-hour limit and uses weekly resets.
How many messages or prompts do I get with Astra on my plan?
OpenAI publishes official 5-hour local-message estimates on its Codex pricing table: 5 to 45 local messages on Plus and Business Standard, earlier published references of 25 to 225 for Pro 5x and 100 to 900 for the previous Pro 20x allowance. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex; those reference figures are not current caps. Pro still has included allowances; check Settings > Usage for remaining allowance and reset times. Business Premium has no 5-hour limit and provides 5x Standard included usage on a weekly reset schedule. These figures are estimated message ranges rather than guaranteed fixed prompt caps; actual usage varies depending on the task, reasoning effort, context size, and whether Fast mode is used. Weekly allowances remain unpublished.
Can I use GPT-6 Astra on ChatGPT Plus?
Yes, within ChatGPT Work and the Codex CLI during rollout. OpenAI provides limited Astra usage on Plus for Work and Codex, while GPT-6 Pro in ordinary ChatGPT web chat is not included with Plus.
What is the minimum Codex CLI version required for Astra?
Codex CLI version 0.153.0 or newer is required to configure and run GPT-6 Astra. You can verify your installed version by running 'codex --version' in your terminal.
What happens if I reach a limit during an active turn?
According to OpenAI documentation, if you reach a usage limit during an active turn, Codex can continue working on that turn, subject to fair-use limits. Once that active turn concludes, subsequent prompts pause until your allowance recovers, credits are added, or an eligible reset is applied.
Does the local Codex Usage tool track Astra usage separately?
No. OpenAI does not expose model-attributed token or turn metrics through its status endpoints. The local Codex Usage tool displays your overall account-level 5-hour and weekly remaining percentages and reset times as reported by the official Codex client, covering all models in your shared pool.
How do I continue coding if Astra exhausts my weekly Codex limit?
If your weekly allowance is exhausted, you can: (1) wait for your scheduled weekly reset, (2) apply a banked reset if one is available on your account, (3) purchase a paid instant reset or add credits in Settings > Usage (available to eligible personal Plus and Pro accounts), or (4) authenticate the Codex CLI with an OpenAI API key to switch to metered OpenAI Platform API billing. A five-hour refresh does not restore an exhausted weekly allowance. Business Premium has no five-hour limit.
- Codex usage limits table — every plan in one table.
- ChatGPT Plus Codex limits — the full Plus breakdown.
- ChatGPT Pro Codex limits guide — the Pro tiers.
- ChatGPT Business Codex limits for Standard vs. Premium seats — how Astra allowances and 5-hour limit rules differ across seat types.
- when your Codex usage resets — recovery timing for both windows.
- Codex credits, paid instant resets, banked resets, and API billing — what to do once Astra has drained your allowance.
- why am I rate limited in Codex — throttled sessions and 429 errors.
- Codex recommended model and Power presets — how Astra fits into the model options in the interface.
- Codex Fast mode speed and credit multipliers — how Fast mode accelerates Astra and affects allowance consumption.