- On your own account (free, Plus or Pro — new subscription, renewal or upgrade): the ChatGPT Pro 200 top-up for ¥1,300 (≈ 193.75 USDT), usually done in 5–30 minutes; your chat history stays put.
- Prefer a separate account, ready to use: a ready-made Pro 200 account for ¥1,349 (≈ 201.05 USDT), already active; hand-delivered after you verify your delivery email (the order page shows 30 minutes to about 4 hours); you can bind your own email.
- Need Ultrafast or the highest cap: Pro 500 on your own account for ¥3,599 or ready-made for ¥3,699.
Codex comes with every plan — the tier decides how much
OpenAI's Help Center is explicit: Codex is included across ChatGPT plans, including Free and Go, and you connect it by signing in with your ChatGPT account in the desktop app (Codex mode), the CLI, the IDE extension or the web. There is no separate Codex subscription. What your tier buys is the allowance: the pricing page lists Pro with “your choice of 3 usage tiers” and “longer Codex and ChatGPT Work sessions”.
Codex on each tier today (30 September 2026)
| Tier | OpenAI price | Five-hour limit in Codex | Allowance vs Plus | Ultrafast |
|---|---|---|---|---|
| Plus | $20 | Yes: an estimated 5–45 GPT-6 Astra local messages per 5 hours | 1× | No |
| Pro 100 (formerly Pro 5x) | $100 | None at the moment; weekly limits apply | OpenAI said 5× in mid-September; no multiple published now | No |
| Pro 200 (formerly Pro 20x) | $200 | None at the moment; weekly limits apply | About 10× for new subscriptions, down from about 20× (Engadget); subscriptions active at OpenAI's cutoff keep the old allowance until 2026-10-29 | No |
| Pro 500 (new) | $500 | None at the moment; weekly limits apply | About 25× (reported, citing OpenAI) | Yes |
The footnotes on OpenAI's pricing page matter more than any multiple. Local messages and cloud chats share the allowance, and cloud tasks can use more of it than local messages. Weekly limits may apply. The estimates are not fixed message limits: model choice, context length, reasoning, tool use and caching all change consumption, so prompt length alone is not a reliable guide — the usage dashboard (or /status in the CLI) is. And GPT-5.5 was retired from Codex, ChatGPT Work and ChatGPT on 14 October 2026, so plan your work around the GPT-6 models.
What the old 5x / 20x table said (24 September)
For readers who arrived looking for those numbers: on 24 September 2026 OpenAI's table estimated GPT-6 Astra at 100–900 local messages per 5 hours on the $200 tier (then Pro 20x), 25–225 on the $100 tier (then Pro 5x) and 5–45 on Plus, with GPT-6 Sol at 300–3,000 / 70–700 / 15–150 and GPT-6 Luna at 7,000–56,000 / 1,750–14,000 / 350–3,000. Those Pro columns are gone: the Pro tiers no longer have a five-hour window, and a new Pro 200 gets a smaller allowance than the one behind the 100–900 figure.
What this means in practice
- On Pro, the weekly limit is the wall. Without a five-hour window a long agent run is not cut off at the five-hour mark; what runs out is the weekly allowance. Pace large refactors across the week and check the dashboard before starting one on a Friday.
- The Codex pool is separate from Chat. OpenAI states that Work and Codex have allowances separate from Chat. Using up your GPT-6 Pro messages in Chat doesn't touch your Codex budget, and vice versa.
- “20x is 4× the $100 tier” no longer holds for new subscribers. Reports put a new Pro 200 at about 10× Plus, against the 5× OpenAI quoted for the $100 tier in mid-September. Only a grandfathered Pro 200 keeps the old allowance, and only until 2026-10-29.
- Ultrafast buys time with allowance. On Pro 500, Ultrafast reportedly reaches about 300 tokens per second in Codex and Work, but uses the included allowance 8× as fast as Standard (Fast mode: 2.5×). Switch it on for the steps you are actually waiting on.
- Astra is what drains it. A refactor that needs forty agent steps in a row, or a cloud task that reads a large repository, spends GPT-6 Astra allowance fast. Routine edits and tests go further on GPT-6 Sol or Luna.
Which developers need which tier
- Occasional Codex, a few hard problems a week: Plus, topped up with Codex credits when you hit the wall. Credits extend usage on your current plan; they are not an upgrade.
- Daily Codex, hitting Plus's five-hour window: Pro 100. No five-hour limit, GPT-6 Pro in chat, and the same features as the higher tiers.
- Agent-first work that empties Pro 100's weekly limits several times a week: Pro 200 — remembering that a new subscription gets less than before the pause. If you already hold a grandfathered Pro 200, keep it running.
- You need Ultrafast, or several agents running for hours: Pro 500. On the reported multiples it is not cheaper per unit than Pro 200; you pay for speed and headroom.
Conversely, if you work in short focused sessions, mostly on Sol or Luna, and rarely see a limit, you don't need to move up. The count-your-interruptions rule in Is Pro 20x worth $200 in 2026? applies to developers exactly as it does to everyone else.
When a subscription is the wrong tool
OpenAI's own pricing page draws the line: for automation in shared environments such as CI, use an API key. You pay per token at API prices, model availability follows your key, and you get no cloud features (GitHub code review, Slack). A subscription — any tier — is for one person working interactively, and OpenAI's terms prohibit sharing an account or using it to power a third-party service. Pipelines belong on the API; your own editor and terminal belong on the subscription.
Getting Pro 200 or Pro 500 now
From 2026-09-10 to 2026-09-30 OpenAI paused new sign-ups and upgrades to the $200 tier, and a ready-made account was the only way in. That is over: the tier is on sale again as Pro 200, and Pro 500 sits above it. Through ClaudeMax there are two routes:
- On your own account: Pro 200 for ¥1,300 (≈ 193.75 USDT) or Pro 500 for ¥3,599 (≈ 536.37 USDT). Free, Plus and Pro accounts all qualify, for a new subscription, a renewal or an upgrade; you submit the sign-in data the order page asks for, it is checked before you pay, and delivery usually takes 5–30 minutes. Your Codex CLI, IDE extension and desktop app keep working with the same sign-in.
- A ready-made account: Pro 200 for ¥1,349 (≈ 201.05 USDT), with the tier already active, for one person; you can bind your own email. After paying you verify a delivery email and our team sends the account (the order page shows 30 minutes to about 4 hours); then sign in to Codex with it like any ChatGPT account.
Payment is USDT on TRC20 or BEP20; contact the bottom-right live chat promptly if anything goes wrong. Prices in detail: ChatGPT Pro 20x (Pro 200) price. Questions: Telegram @ccnb1937. ClaudeMax is an independent reseller operating since November 2024, not OpenAI.
In one sentence
Codex comes with every plan, the Pro tiers currently have no five-hour limit in Codex, a new Pro 200 gets roughly half the Codex allowance the old 20x had according to reports, Pro 500 adds Ultrafast at 8× the allowance cost, and CI belongs on an API key.