cocodot
← Back to guides
Local card declined for overseas AI? cocodot: one card + one key
ChatGPTUpdated 2026-10

ChatGPT Pro at $200 or $500 a Month: What Each Tier Actually Buys, and Whether You Need It (2026)

ChatGPT Pro has a $200 tier and, since DevDay 2026, a $500 one; the difference from Plus is allowance and speed rather than answer quality. What the $500 tier adds, how to compute whether either tier pays for you, the option most people miss, and what helps a large recurring charge go through.

TL;DR: ChatGPT Pro is the $200/month tier, ten times the price of Plus, and the difference is one thing: allowance. Pro gives you a much higher allowance on the strongest reasoning model, high allowances on heavy features like video and deep research, and early access to new capabilities — not a smarter model. On the same question, the answer-quality gap between Plus and Pro is nowhere near as large as the price gap. The decision rule is simple: if you spend time each week stopped and waiting because you hit the Plus ceiling, multiply that time by what your hour is worth; if it exceeds the $180 monthly difference, upgrade. If you do not hit the ceiling, do not. There is also a route people overlook — if what you need is your own program calling a model, rather than working in the chat box or in Codex, you want metered API billing, not any subscription tier. On payment, a large recurring charge is scored far more tightly than $20. October 2026 update: at DevDay 2026 (September 29) OpenAI introduced a new $500/month Pro tier with the highest usage limits and access to GPT-6 Astra Ultrafast (in ChatGPT, available on Pro $500 and eligible Enterprise and Edu plans), which OpenAI says runs up to 8x faster than standard Astra in ChatGPT Work and Codex. GPT-6.1 Sol, meanwhile, is open to both Plus and Pro in ChatGPT Work and Codex, though not yet in chat.

October 2026 update: after DevDay, Pro has a $500 tier

At DevDay 2026 on September 29, OpenAI introduced a new $500/month Pro tier, which it frames as meeting higher building demands. In OpenAI's own words, the tier offers the highest usage limits and access to GPT-6 Astra Ultrafast, which OpenAI calls its fastest frontier model, running up to 8x faster than standard Astra in ChatGPT Work and Codex. The 300-tokens-per-second figure in the DevDay recap is given for Codex: up to 8x faster token generation in Codex (300 tokens per second). On the ChatGPT side, Astra Ultrafast is available on Pro $500 and eligible Enterprise and Edu plans. Separately, in its Agents API announcement OpenAI says $500 Pro subscribers can also use computer use in Codex and ChatGPT Work. One detail matters for the arithmetic: OpenAI's ChatGPT help documentation (learn.chatgpt.com) says that on Pro $500, Ultrafast uses your included usage first and your credits after that, and that for GPT-6 Astra it consumes subscription limits at 8x the Standard rate. Running with Ultrafast on burns allowance much faster — speed has a price. When does the $500 tier make sense? If you still hit limits regularly on the $200 tier, or your work is genuinely bottlenecked on waiting for generation — watching Codex grind through long runs — then yes. If you only run short occasionally, the formula in section 2 still applies; just use the gap between the two tiers instead. The tiers and prices actually available to your account are whatever OpenAI's own subscription page shows. The main DevDay announcements are summarised at /hub/openai-devday-2026.

October 2026 update: dots and GPT-6.1 Sol — what Plus and Pro each get

Two more DevDay announcements bear directly on which tier you pick. dots: OpenAI describes them as powered by GPT-6 Astra, each with its own cloud computer, able to work across the apps you choose to connect — available on Pro, Business Premium and Enterprise. Plus is not on that list, and the DevDay recap says "Pro" without naming a price tier. GPT-6.1 Sol: OpenAI says it brings Sol closer to GPT-6 Astra on hard tasks at one fifth of Astra's standard input and output token prices. It is available from launch to all Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, but in OpenAI's words it is not yet available in chat — so do not look for it in the ordinary chat box. The same ChatGPT help documentation adds two details: in Codex it is the desktop app and CLI, in ChatGPT Work it is web and mobile; and Ultrafast for GPT-6.1 Sol is coming later. The practical upshot: if all you want is access to GPT-6.1 Sol, Plus already has it (allowance aside); dots need Pro; Astra Ultrafast needs the $500 tier.

1. What Pro adds over Plus — and what it does not

Clear up the easiest misconception first: Pro buys allowance, not intelligence. Within the same generation there is no smarter edition behind a higher price, and the Astra Ultrafast announced at DevDay is a faster mode of the same GPT-6 Astra — it buys speed, not a higher level of intelligence. Pro's value sits in four places: the strongest reasoning model comes with a much higher allowance, where Plus has explicit limits by count or time window; heavy features such as video generation and deep research come with allowances an order of magnitude larger; new capabilities usually open to Pro first; and since DevDay some capabilities are gated by tier — dots are available on Pro, Business Premium and Enterprise, Astra Ultrafast, on the ChatGPT side, only on the $500 Pro tier and eligible Enterprise and Edu plans, and the ChatGPT help documentation notes that some paid plans omit Astra's Extra High reasoning preset. So "are Pro's answers better?" is the wrong question. The right ones are "will I ever consume that much allowance?" and "do I need any of the tier-gated features?"

Open a cocodot US-BIN virtual card for overseas AI subscriptions; users have paid for ChatGPT with it. Start with a small first payment.

$9.9 card fee (one-time) · 3% top-up fee · $1 monthly fee per active card · top up via Alipay / WeChat Pay (RMB), USDT / USDC

2. A formula you can actually compute

Do not decide on feel. The gap is $180 a month. Use this: (minutes per week spent stopped and waiting on a rate limit × 4 weeks ÷ 60) × your hourly rate. If that figure exceeds $180, upgrading pays; otherwise it is waste. Concretely: someone billing $30 an hour who loses three hours a week to limits is losing twelve hours, around $360 a month — comfortably more than the difference, so upgrade. Someone blocked for twenty minutes a week loses about $40 a month and would be paying $180 to recover it. Most people agonising over this question do not hit the ceiling more than a handful of times in a month, which is itself the answer.

3. Measure your real usage before you guess

Most people skip this step, and it is the most valuable one. Think back over the last fortnight: on how many days did you actually see a limit message? Never — your current tier is sufficient and upgrading only wastes money. Three or more times a week — you are genuinely a heavy user, read on. There is a sharper angle too: notice whether your usage clusters into a handful of days. A very common pattern is three or four frantic days on a deadline and barely any use for the remaining twenty-odd, with the tier chosen on the experience of those few days. For that shape, a subscription is always the worst way to pay — you are funding three weeks of idleness.

4. The counter-intuitive third option: perhaps do not subscribe

This one saves people four figures. Subscriptions are designed for use in a browser or app. If you mainly call from code — writing scripts, building agents, processing documents in bulk, running automations — then a web subscription is doubly wasteful: the allowance cannot be reached from your program, and metered billing is much cheaper for lumpy usage anyway. The test is one sentence: chat box, or code? A common optimum is in fact the cheapest subscription plus metered API — the subscription for everyday chat, where the app and the convenience matter, and the API for code and long-running tasks, where metered billing and programmatic access matter. Total spend usually lands well below the Pro tier.

October 2026 update: if you write code, separate Codex from the API

The previous section says "code means API". After DevDay that needs one more cut, because Codex itself is part of the subscription. Having AI write code for you — editing code, running tasks and reviewing PRs in Codex — draws on subscription allowance: GPT-6.1 Sol is open to Plus, Pro, Business, Enterprise and Edu; Codex Cloud (run on your computer, steer from your phone, or run in the cloud) and code review on GitHub PRs and GitLab MRs are, in OpenAI's words, available on all plans; and the ChatGPT help documentation states that ChatGPT Work and Codex share the same usage. For this kind of work a subscription is the right tool, and you upgrade only when you genuinely hit limits. Having your own program call a model — scripts, agents, batch jobs, a product you ship to customers — is API work, and subscription allowance cannot reach it. On OpenAI's API, GPT-6.1 Sol is gpt-6.1-sol at a standard price of $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens; worked cost examples and payment troubleshooting are at /hub/gpt-61-sol-api-china. From China there are two routes. First, use OpenAI's API directly and fund it with a card on the OpenAI platform — locally issued cards are often declined at this step, and a US-BIN virtual card solves it (see /card); cocodot's virtual cards have had successful charges on OpenAI API top-ups. Second, use cocodot's OpenAI-compatible relay: top up with Alipay, change the base_url, no card needed. To be clear, the relay does not carry the GPT-6 family yet (Astra, Sol, 6.1 Sol, Luna); it serves GPT-5.6 Sol / Terra / Luna, GPT-5.5, Claude, Gemini and others, with live prices at cocodot.co/pricing and the GPT-5.6 differences at /hub/gpt-56-sol-terra-luna-api. If you need GPT-6.1 Sol, take the first route; if GPT-5.6 or Claude is enough, the second is simpler.

5. The real difficulty with Pro: large recurring charges

Payment, not access, is the obstacle, and a $200 monthly charge is not merely a bigger version of a $20 one — the larger the amount, the tighter the risk scoring on both the issuer's and the merchant's side, and recurring large charges attract particular attention. Three things to check (whether a payment goes through is up to the merchant's review). ① The BIN — platforms read the issuing institution's country from the first digits of the card number, so a locally issued Visa is still scored as a local card; the issuing country has to change, not the network. ② The billing address — the card is not tied to a personal address; enter a complete address that is genuinely yours, keep the same one for the same merchant account, and never use an address found online, someone else's, or a made-up one. ③ The balance — $200 plus headroom for FX movement must be on the card on the renewal date, or the second month simply stops. If the tier you want is the $500 one, the same three rules apply unchanged — the amount has simply more than doubled again, and scoring only gets more sensitive. On renewal day the card needs $500 plus headroom for FX and temporary holds, not exactly $500.

6. Paying with a virtual card, end to end

Paying on the official site yourself, without handing your account to anyone: ① sign up at cocodot and top up the wallet, Alipay included, with current fees published at cocodot.co/pricing; ② issue a US-BIN virtual card (see /card for choosing a BIN) and move balance onto it — wallet balance is not card balance, and you want above $210 rather than exactly $200 (for the $500 tier, budget $500 plus the same kind of headroom), leaving room for FX movement and temporary authorisation holds; ③ in ChatGPT open Settings → Subscription, select Pro (check which price tier you are choosing), and enter the card details with a complete billing address that is genuinely yours (the card is not tied to a personal address), keeping the same one for the same merchant account — never one found online, someone else's, or a made-up one; ④ once the first month succeeds, note the renewal date and confirm the card balance two or three days ahead every month. The same card has also had successful charges on Claude, OpenAI API top-ups and OpenRouter, so there is no need for a separate card per platform.

7. What to do after a decline — the most important section

The worst response to a declined large subscription charge is retrying repeatedly. Several failures in a short window get the account marked as risky, after which even a completely sound card may keep being refused — this is where people break an account that would otherwise have worked. The correct sequence: stop after two or three failures, confirm the BIN, the billing address and the balance one at a time, and try again twenty-four hours later. If checkout asks for 3-D Secure, that is not an error but the normal risk flow — complete it. One more thing worth knowing: starting on Plus for a full month before deciding to upgrade is the cheapest possible way to find out; which plan changes are available, and how they are settled, is whatever your account's subscription settings page shows.

Plus, Pro and the API (October 2026 update: adds the $500 tier from DevDay)

Plus $20/moPro $200/moPro $500/mo (new at DevDay)Metered API
SuitsAlmost everyone using the chat boxRunning heavy tasks continuously, every dayOpenAI's framing: higher building demandsAnyone whose own program calls a model
AllowanceCapped; heavy users reach itMuch higher than PlusOpenAI says: the highest usage limitsNo subscription allowance — you pay for what you use
Hitting the limitWait for the window to refreshRarely happensExact limits not publishedNo allowance window; request-rate and concurrency limits still apply
SpeedStandardStandardGPT-6 Astra Ultrafast (Pro $500 and eligible Enterprise and Edu): up to 8x standard Astra in ChatGPT Work and Codex (draws allowance at 8x the standard rate)Ultrafast up to 6x faster in the API
New at DevDayGPT-6.1 Sol (ChatGPT Work and Codex)GPT-6.1 Sol; dots (OpenAI says "Pro" without naming a price tier)GPT-6.1 Sol; dots; computer use in Codex and ChatGPT Workgpt-6.1-sol and Astra Ultrafast at API rates
When usage is lumpyFixed cost, wasted in quiet monthsFixed cost, more wastedFixed cost, the most wastedA quiet month costs less automatically
Callable from a programNoNoNoYes — that is what it is for
Paying for itOverseas card, $20/moOverseas card, $200/mo, scored harderOverseas card, $500/mo, the largest amount, scored hardestCard on Billing, or a relay taking local payment

FAQ

Is Pro's model smarter than Plus's?

No. Within a generation both tiers reach the same model, and the Astra Ultrafast announced at DevDay is a faster mode of the same Astra. Pro buys allowance — a much higher allowance on the strongest reasoning model, larger allowances for video and deep research, early access to new features — plus tier-gated features such as dots. If you have never hit a Plus limit and do not need those features, upgrading will not improve your answers.

Is $200 a month worth it?

Compute it: weekly minutes lost waiting on rate limits, times four weeks, times your hourly rate. If that exceeds the $180 difference, it is worth it. Heavy users blocked more than about three hours a week usually come out ahead; anyone hitting the ceiling a few times a month does not.

What is the difference between the $500 and $200 Pro tiers?

Per OpenAI, the $500 tier carries the highest usage limits and access to GPT-6 Astra Ultrafast (on the ChatGPT side, Pro $500 and eligible Enterprise and Edu plans), up to 8x faster than standard Astra in ChatGPT Work and Codex; $500 Pro subscribers can also use computer use in Codex and ChatGPT Work. Note that Ultrafast draws on your allowance at 8x the standard rate. If you are only short of allowance, look at the $200 tier first; the $500 tier makes sense when generation speed itself is what holds your work up. The tiers and prices actually on offer are whatever OpenAI's own subscription page shows.

Can I call GPT-6.1 Sol through cocodot's API?

Not at the moment. cocodot's OpenAI-compatible relay does not carry the GPT-6 family yet (Astra, Sol, 6.1 Sol, Luna); what it serves includes GPT-5.6 Sol / Terra / Luna, GPT-5.5, Claude and Gemini, with live prices at cocodot.co/pricing. For GPT-6.1 Sol, use OpenAI's own API (model name gpt-6.1-sol) and fund it with a US-BIN virtual card; cocodot's virtual cards have had successful charges on OpenAI API top-ups.

I mostly write code — should I buy Pro?

Separate two things. Having AI write code for you — editing code, running tasks and reviewing PRs in Codex — draws on your subscription allowance, and GPT-6.1 Sol is open to both Plus and Pro there, so start on Plus and upgrade only if you genuinely hit the limit. Having your own program call a model — scripts, agents, batch jobs — cannot use subscription allowance at all; that is metered API territory, and buying Pro does not help.

Why is $200 harder to charge than $20?

Risk thresholds tighten as amounts grow, and recurring large charges are watched especially closely. What helps is a US-BIN card rather than a local one, a complete billing address that is genuinely yours and kept the same for that merchant account, and a balance comfortably above the fee rather than exactly equal to it.

Can I switch back to Plus later?

Whether you can switch, when it takes effect and how it is settled follow the options on your ChatGPT account's subscription settings page. The cheapest way to test is still a full month on Plus first, upgrading only once you have confirmed the allowance genuinely is not enough.

What happens if the card is short on the renewal date?

The charge fails and the subscription lapses. Prepaid-card users should note that wallet balance is not card balance — move the money across in advance and leave headroom for FX movement, budgeting above $210 rather than exactly $200.

About cocodot

cocodot is a payment and AI access service for developers and cross-border teams in mainland China. It provides US-BIN virtual cards issued by a licensed institution — used to pay for overseas subscriptions and ad accounts — and an OpenAI-compatible AI API gateway for calling Claude, GPT and Gemini from within mainland China. Both share one wallet, funded by Alipay / WeChat Pay (RMB), USDT / USDC and accounted in USD. Card: $9.9 to open, 3% to load, $1 per active card per month; spending: $0.60 settlement fee on purchases under $20; a corresponding fee applies when the issuer charges one.

Service scope, pricing and limits →
US-BIN virtual card for overseas AI subscriptionsOpen a cocodot card