GPT-5.5 retired on October 14, 2026: what to switch to in ChatGPT, Work and Codex, what to change, and which plan is enough

When did GPT-5.5 retire, and what is affected? (checked 2026-10-11, rechecked 2026-10-14)

All of this comes from the GPT-5.5 retirement notice at the top of the Models page in OpenAI’s Codex documentation; the pricing page carries the same notice, and the model card on the Models page files GPT-5.5 under previous-generation flagship. Two things OpenAI does not say: what a client does after the retirement if it still names gpt-5.5 (an error, or a silent switch to another model), and how existing Chat conversations on GPT-5.5 are carried over. So if anything still names it, change it now.

What replaces GPT-5.5 in ChatGPT, ChatGPT Work and Codex?

Chat already moved on October 7: GPT-6 started rolling out to paid plans that day and to Free and Go from October 8, and anyone on Latest was carried to GPT-6 at the same thinking level. What each plan gets in Chat is covered in GPT-6 and Intelligent UI in ChatGPT. For which of Sol, Luna and Astra suits which job in Work and Codex, see GPT-6 Sol and Luna in Codex and Work, and for the developer’s view of the Pro tiers, ChatGPT Pro for developers.

How to replace gpt-5.5 in Codex: the migration checklist (7 steps)

OpenAI’s instruction is to replace a hard-coded gpt-5.5 in 6 places: workspace defaults, saved model settings, managed configurations, custom agents, scheduled tasks and scripts that select a model. The steps below run from what an individual controls to what an administrator does.

1. Check how you sign in: The retirement applies to ChatGPT, ChatGPT Work and Codex when you sign in with a ChatGPT account; GPT-5.5 has been gone there since October 14, so if you have not switched yet, do it now. Calls to the OpenAI API need no change, and Codex signed in with an API key follows the models available to that key, which this retirement does not touch.
2. Pick the replacement: On Plus, Pro, Business, Enterprise and Edu choose GPT-6 Sol (gpt-6-sol). On Free and Go choose GPT-6 Luna (gpt-6-luna) in the desktop app. OpenAI marks both as available when they reach your account, so go by your own model menu.
3. Change your default and saved model settings: In the desktop app, ChatGPT Work on the web and the IDE extension, use the model and reasoning control beneath the composer; in the Codex CLI use /model. If config.toml has a model entry set to gpt-5.5, change it to gpt-6-sol. If you never set a model, leave it alone: the client uses a recommended one.
4. Search your scripts and CI: Search scripts, CI configuration and shell aliases for gpt-5.5. Anything like codex --model gpt-5.5 or codex exec -m gpt-5.5 should name gpt-6-sol instead.
5. Check custom agents and scheduled tasks: Both are on OpenAI’s list. Open each one and replace a pinned gpt-5.5. Scheduled tasks run unattended, so they are the easiest to miss.
6. Admins: workspace defaults and managed configurations: On Business, Enterprise and Edu, check the workspace default model and any managed configuration you distribute. Selecting a new model does not grant access to it: in Enterprise and Edu, GPT-6.1 Sol stays off until an administrator enables it.
7. Run a familiar task, then check usage: OpenAI notes that reasoning efforts do not map exactly between model generations: try a familiar task at a lower setting and adjust from the result. Then open Settings → Usage (or run /status in the Codex CLI) to see how your allowance is being used.

If Codex says the model is not available (from October 14): run /model in the CLI, pick a model your account has, and confirm it works; then go back through config.toml and the --model flags in your scripts until no gpt-5.5 is left. If the replacement is missing from the menu, update the client first (for GPT-6 Astra, OpenAI requires Codex CLI 0.153.0 or newer); on Enterprise and Edu, ask your administrator about model access. OpenAI has not published the exact error text, and we will not guess at it.

Does GPT-6 Sol cost more than GPT-5.5? Credit rates compared

The short answer: not by the rates. The official replacement is charged fewer credits per token than GPT-5.5 was. The table uses the credit rates in OpenAI’s pricing documentation on 2026-10-11 (Standard speed, credits per 1M tokens); the multiples and the last numeric column are our arithmetic on those rates.

• GPT-6 Sol. Input and cached input at 40% of the GPT-5.5 rate, output at 33%. Counting output only, one 500-credit pack goes from about 0.67M tokens to about 2M.
• GPT-6.1 Sol. Same input and output rates as GPT-6 Sol, with a lower cached-input rate: 20% of GPT-5.5’s.
• GPT-6 Luna. Input at 2% of the GPT-5.5 rate. It is built for focused, high-volume tasks; do not expect it to stand in for GPT-5.5 on complex work.
• GPT-6 Astra. The only model here that costs more than GPT-5.5: 2x on input and about 1.67x on output. Do not swap your default from GPT-5.5 to Astra out of habit, or your allowance will drain faster.

Four cautions when reading it:

1. These are Standard-speed rates. Fast mode spends credits at 2x and included usage at 2.5x; Ultrafast is 6x and 8x, and only on Pro 500.
2. The rates govern how credits are spent. OpenAI does not publish included plan usage from this table and says credit prices alone do not determine it; your usage dashboard is the authority.
3. Cheaper per token is not the same as cheaper per task. The same job may use a different number of tokens on another model, and reasoning efforts do not map one to one. Run a task you know and watch what it uses.
4. API token prices are a separate price list; OpenAI says not to use them to estimate subscription usage.

What each plan gets in Codex and Work since October 14

Three notes on reading it. First, the five-hour row is OpenAI’s estimate for Plus (a range such as 15–150 local messages on GPT-6 Sol), not a fixed count; local messages and cloud tasks draw on the same allowance, and cloud tasks may use more of it. Second, OpenAI’s own wording for Pro is that there is currently no five-hour limit; weekly limits may still apply, and Settings → Usage shows what your account has. Third, the pricing page gives Free and Go GPT-6 Luna at Standard speed in the desktop app only, subject to rollout, and lists GPT-6.1 Sol and GPT-6 Luna on the Plus card.

Do I need a different plan because GPT-5.5 has retired?

The honest answer: for most people, no. This is a model change, not an allowance change. The replacement is charged fewer credits per token; Plus’s five-hour estimates are row for row what they were when we checked on 2026-10-08, and Pro plans still have no five-hour limit (checked 2026-10-11). Change the setting, work for a couple of days, and if the usage dashboard looks normal, you are done. The signals that do justify a different plan are about how often you hit a limit:

Where credits stop being the cheaper route (today’s prices): Plus with up to 3 credit packs a month still costs less than a Pro 100 top-up (108.05 USDT against 126.68 USDT); Pro 100 with up to 2 credit packs a month still costs less than a Pro 200 top-up (180.33 USDT against 193.75 USDT); Pro 200 with up to 12 credit packs a month still costs less than a Pro 500 top-up (515.65 USDT against 536.37 USDT). This is a price comparison only: credits cover eligible usage in Codex and ChatGPT Work, they do not add GPT-6 Pro messages in Chat, and they do not unlock Ultrafast. The longer version is in Hit your Codex usage limit? and Buy ChatGPT Codex credits.

Do not buy when:

• The retirement notice is the only reason. Changing a model setting costs nothing.
• You hope a higher tier brings a model sooner. It does not. OpenAI marks the replacements as available when they reach your account, and on Enterprise and Edu an administrator controls access.
• You want Ultrafast but plan to add credits on Pro 100 or Pro 200. That does not unlock it; Ultrafast is on Pro 500 only.
• You plan to split one Pro account between several people. Don’t. OpenAI’s terms prohibit sharing credentials and reselling access, and shared use is how accounts get restricted.

If the real obstacle is that chatgpt.com declines your card or does not serve your country, that is the part we solve: you pay in USDT (TRC20 or BEP20) and the plan or the credits land on your own account.

Pro 100, 200 or 500: what one unit of usage costs

The three tiers share the same chat model and features; they differ in allowance and speed. On allowance, the Help Center says only that Pro 200 includes more usage than Pro 100 and Pro 500 the most. Divide each monthly price by the multiple reported in the press (Codex and Work usage relative to Plus, not official figures) and you get the price of one Plus-worth of usage:

How to read it: at OpenAI’s prices every tier costs about $20 per unit, so the differences come from our pricing. At our prices a unit on Pro 100 costs about 25.34 USDT against 27.58 USDT on Plus, so if Plus runs out most days, Pro 100 is the better deal. Pro 200 costs 1.53x as much as Pro 100 for about 2x the usage (reported), so each unit is cheaper on Pro 200: 19.38 USDT against 25.34 USDT. At today’s prices the cheapest unit on our list is Pro 200 (new allowance), about 19.38 USDT. The multiples are rough ratios for Codex and Work, and Settings → Usage is the authority for your own account.

Against OpenAI’s list price. Taking 1 USDT ≈ 1 USD, and before any foreign-card fees or taxes on OpenAI’s side: Pro 100 is 126.68 USDT here, 26.68 USDT more than OpenAI’s $100; Pro 200 is 193.75 USDT here, 6.25 USDT less than OpenAI’s $200; Pro 500 is 536.37 USDT here, 36.37 USDT more than OpenAI’s $500. Where we cost more, the difference pays for convenience: no card that chatgpt.com has to accept, payment in USDT, and a check before you pay.

More detail: ChatGPT Plus vs Pro 100, ChatGPT Pro 100 vs 200 vs 500, Pro 200 vs Pro 500, and the monthly ChatGPT Pro price list.

Buying here: top-up, ready-made account, or Codex credits

In one line: keep your history with a top-up; take a ready-made account if you would rather not submit sign-in data or want a separate work account; buy credits if you only overflow now and then. Pro is monthly-only at OpenAI. Do not keep auto-renewal on at OpenAI or the App Store at the same time, or you may be charged twice. A ready-made account is a different account, so chats and settings saved on your old one do not follow; when you set it up, choose GPT-6 Sol from the start. The ordering details per tier: How to buy ChatGPT Pro 100, Pro 200 and Pro 500.

What we guarantee, and what we cannot promise

What we guarantee in writing:

1. Checked before you pay. A top-up checks your sign-in data first, and you pay nothing if it fails; the data is used once and cleared after delivery.
2. One ready-made account, one buyer. Each ready-made account goes to a single buyer, and the order page shows its status at every step.
3. Refund terms are public. 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. Contact the bottom-right live chat promptly if anything goes wrong.
4. So are the exclusions. Bans caused by breaking OpenAI’s usage policy are not refunded.
5. Track record. ClaudeMax has operated since November 2024 and has served more than 20,000 users (our own figure).
6. You can verify delivery yourself. Settings → My Plan shows your Pro tier on OpenAI’s own interface.

What we cannot promise:

• We are not OpenAI. ClaudeMax is independent: not OpenAI, and not an authorized OpenAI reseller.
• OpenAI decides which models stay and how much each plan includes. When a model retires, when a replacement reaches a given account, how much each plan includes and whether that changes again are OpenAI’s decisions; no upgrade keeps GPT-5.5 around.
• OpenAI’s Terms of Use. Sharing account credentials, reselling access and automatically extracting data are prohibited, and OpenAI may restrict access for misuse. We cannot change that; we can help you keep the risk low: one account, one person; change a ready-made account’s password at once; never lend or split it; avoid signing in from many different countries.
• Prices are not fixed forever. They follow cost, exchange rate and stock; the pricing page is authoritative.

Want to check us out first? See Is ClaudeMax legit?