ChatGPT Pro 200 for coding (October 2026): Codex limits, models, Ultrafast and who actually needs it

ChatGPT Pro 200 in Codex vs Plus, Pro 100 and Pro 500

Every paid plan gets the same Codex: the desktop app, CLI, IDE extension, web, iOS, Codex Cloud and the GitHub, Slack and Linear integrations, with the same models. What changes from tier to tier is how much you can run before a limit, plus a few Pro-only extras. Facts below are from OpenAI's pricing docs and Help Center, read 2026-10-20; usage multiples are press reports and labelled as such.

About the allowance. A Pro 200 bought now starts on the new, lower allowance (reported at about 10x Plus in Codex and Work), unless the account had an active Pro 200 between September 22 and September 29; from October 30 everyone is on it. OpenAI's Help Center no longer publishes per-plan numbers; it says only that Pro 200 has more usage than Pro 100 and Pro 500 the most. Your own Settings → Usage page shows the real limit and reset time.

How tight Plus is, for comparison

OpenAI estimates Plus at these local Codex messages per five hours (not a hard cap: bigger tasks mean fewer messages, and cloud tasks use more). OpenAI describes Plus as enough to "power a few focused coding sessions each week".

The squeeze is GPT-6 Astra: 5–45 messages disappear quickly on a hard bug. That is the most common reason developers move off Plus.

Which model Codex uses on Pro 200

• GPT-6.1 Sol for most coding. OpenAI recommends it for complex coding and agentic work: "near-Astra performance" at a lower cost than Astra. Set it with /model, codex --model gpt-6.1-sol or model = "gpt-6.1-sol" in config.toml.
• GPT-6 Astra for the hardest end-to-end work. OpenAI notes that Astra at Low effort can beat Sol at High, so try it at a low setting first. It needs Codex CLI 0.153.0 or newer.
• GPT-6 Luna for focused, repeatable tasks like extraction, transformations and quick edits; it uses far less allowance.
• GPT-5.5 is gone. It left Codex for ChatGPT sign-in on October 14. Replace gpt-5.5 with gpt-6.1-sol in saved settings, custom agents, scheduled tasks and scripts.
• Reasoning from Light to Ultra. Max gives one task more thinking time; Ultra splits a task across subagents. Both use more allowance, and on Plus that drains the five-hour window fast, which is one reason Ultra pairs well with a Pro plan.

How much each model costs once you are on credits, per OpenAI's rate card, with our three example tasks (our own assumptions: small = a bug fix, medium = a multi-file feature with tests, large = a long refactor session):

So one 500-credit pack is about 52 medium tasks on GPT-6.1 Sol, but only about 9 on Astra. OpenAI's own reference point: a typical GPT-5.6 Sol task uses 2–15 credits.

Fast and Ultrafast: what speed costs on each plan

• Fast is on every paid plan for GPT-6.1 Sol, Astra, GPT-6 Sol and Luna (/fast in the CLI). It uses included usage at 2.5x the Standard rate and purchased credits at 2x. On Pro 200 that is affordable for interactive work; on Plus it shortens the five-hour window by the same factor.
• Ultrafast covers GPT-6 Astra and GPT-6.1 Sol, and OpenAI says Astra Ultrafast generates tokens up to 8x faster than Standard in Codex. Among Pro plans it is only on Pro 500, at 8x included usage, then 6x credits. Credits bought on Pro 200 do not unlock it.

If you want Ultrafast every day, Pro 500 (¥3,599 ≈ 536.37 USDT) is the only route; see Pro 200 vs Pro 500.

Codex Cloud, Composer predictions and the CLI on Pro 200

• Codex Cloud. Cloud tasks run in a 4 vCPU / 16 GiB / 32 GiB VM on Pro (Plus: 2 / 8 / 8). Cloud and local work share one allowance, and cloud tasks may use more of it, so long cloud jobs are where the missing five-hour window matters most. Setup is in our Codex Cloud guide.
• Composer predictions (beta). After Codex replies, the desktop app suggests your next message; press Tab to take it as an editable draft. Personal Pro only, 18+, local and SSH threads, GPT-6 Astra or GPT-6.1 Sol. Generating predictions does not count toward your limits; toggle it under Settings → General → Composer → Show predictions.
• CLI and IDE. Sign in with the ChatGPT account that has Pro 200 (codex login); /status shows your remaining allowance. Signing in with an API key instead bills API rates and drops the cloud features.
• GPT-6 Pro in Chat. Separate from Codex: its own weekly limit on Pro 200, and since October 16 it can run several agents in parallel. More in How to get GPT-6 Pro.

Which plan for which developer

These profiles are our own grouping, not OpenAI tiers. Monthly cost is at today's prices here (refreshed hourly; the pricing page is authoritative).

A team is different: each person needs their own account, because OpenAI forbids sharing one. For that, ChatGPT Business seats (¥220 ≈ 32.79 USDT standard, ≈ 163.94 USDT premium per seat per month, at least 2 seats) give everyone Codex; by our reckoning a premium seat is about 5x Plus in Codex, with no five-hour limit. See Pro 200 vs Business.

The break-even, computed from today's prices

• Plus + credits vs Pro 100: cheaper up to 3 packs a month (1,500 credits, ≈ 156 medium GPT-6.1 Sol tasks); from 4 packs, Pro 100 wins.
• Plus + credits vs Pro 200: cheaper up to 6 packs (≈ 312 medium tasks); from 7, Pro 200 wins.
• Pro 100 + credits vs Pro 200: cheaper up to 2 extra packs; from 3, move to Pro 200.
• Pro 200 + credits vs Pro 500: cheaper up to 12 extra packs, unless you need Ultrafast, which credits cannot buy.
• Per unit of usage, dividing our top-up price by the reported multiple: Plus ≈ 27.58 USDT, Pro 100 ≈ 25.34 USDT, Pro 200 ≈ 19.38 USDT, Pro 500 ≈ 21.46 USDT per Plus-worth. Today the cheapest unit is Pro 200, provided you actually use the headroom.

Treat this as an estimate. Credits are not the same unit as included usage, the medium task is our assumption (9.6 credits on GPT-6.1 Sol; a small one is about 2.6), and credits never remove the Plus five-hour window: they only let you continue once it is used up. They also last 12 months, so lumpy months cost less than the table suggests. Per-model detail: How far do 500 Codex credits go?

When the weekly limit runs out on Pro 200

1. Check which limit and when it resets: Settings → Usage, the Codex usage dashboard, or /status. If you hit it mid-task, OpenAI lets the agent finish the current turn.
2. Use a lighter model for the rest of the week. Switching does not refill the shared pool, but GPT-6.1 Sol or Luna use far less of what is left than Astra at high effort.
3. Buy credits. Plus and Pro users can keep working on credits without changing plan. Ours: ¥180 ≈ 26.83 USDT per 500 credits, added to your own account.
4. A purchased reset, if OpenAI offers one to your account (eligible personal Plus and Pro, depending on billing country), refreshes the allowance immediately.
5. Every week? Then Pro 500 is the next step. OpenAI Support cannot reset limits.

Getting Pro 200: the top-up keeps your Codex history

The Pro 200 top-up is added to your own ChatGPT account, so nothing in Codex moves: threads, cloud environments, the GitHub connection, config.toml, custom agents and scheduled tasks all stay, as do your chats and memory. Free, Plus and Pro accounts qualify, so an active Plus or Pro 100 upgrades in place. Price: ¥1,300 ≈ 193.75 USDT.

1. Sign in to the account you code with: Open chatgpt.com with the same account you use in Codex and check Settings → My Plan. Free, Plus and Pro accounts all qualify for the Pro 200 top-up (new subscription, upgrade or renewal).
2. Pick the Pro 200 top-up: On the pricing page choose ChatGPT Pro 200 as a top-up on your own account and check the live price and stock.
3. Submit your session, checked first: Paste the session JSON the order page asks for. It is checked before you pay; if the check fails you pay nothing. Your password stays with you.
4. Pay in USDT: TRC20 or BEP20, exactly the amount shown on the order page. Each order covers 30 days.
5. Wait for activation: Usually 5 to 30 minutes, on the same account. The session data is used for this top-up only and cleared after delivery.
6. Confirm it in Codex: Open the usage dashboard or run /status in the Codex CLI: the Plus five-hour window is gone. If the CLI still shows the old plan, run codex logout and codex login. Your Codex threads, cloud environments and GitHub connection are unchanged, because it is the same account.

No ChatGPT account yet, or want a separate work account? A ready-made Pro 200 (¥1,349 ≈ 201.05 USDT, ≈ 7.31 USDT more than the top-up in the first month): pay, enter a delivery email, click the Cloudflare verification email, and our team delivers the account by hand in 30 minutes to about 4 hours. One account per order, for one user. To renew, message Telegram @ccnb1937 with your order ID and pay the top-up price. It starts empty: OpenAI does not move Codex history between accounts. See top-up or ready-made?

Paying in USDT means no card has to pass at chatgpt.com; exact amounts per network are in Buy ChatGPT Pro 200 with crypto, and local-currency equivalents in ChatGPT Pro 200 price in your currency.

Pro 200 or Claude Max 20x for coding?

Both are $200 at list price. Pick by agent, not by benchmark: Pro 200 is Codex with no five-hour window, GPT-6 Astra and GPT-6.1 Sol, Codex Cloud and Composer predictions; Claude Max 20x is Claude Code, whose five-hour and weekly limits are shared with the Claude app. The full side-by-side, with our prices for both, is in ChatGPT Pro 200 vs Claude Max 20x.

Can you trust us? What is in writing

• Checked before you pay. A top-up checks your session first, and you pay nothing if it fails; the data is used once and cleared after delivery.
• Refund terms. If an OpenAI regional policy change makes the service unavailable, you get a refund for the unused days, processed within 1 business day and returned in USDT. Bans caused by breaking OpenAI's usage policy are not refunded. Contact the bottom-right live chat promptly if anything goes wrong.
• Who we are. ClaudeMax has operated since November 2024 and has served more than 20,000 users (our own figure). We are not OpenAI and not an authorized OpenAI reseller; OpenAI sets each plan's Codex allowance.
• One person per account. OpenAI's terms prohibit sharing credentials and reselling access, and OpenAI may restrict access. OpenAI serves only supported countries. More in Is ClaudeMax legit? and the Pro 200 FAQ.