ChatGPT Pro now has three tiers: 100, 200 and 500. If you mainly use Codex, choose by separating the regular usage allowance from the value of faster processing. With current Plus set to 1, the announced regular allowances are 5 for 100, 10 for the new 200 and 25 for 500. Two 100 subscriptions and one new 200 provide the same combined allowance on that scale. In the yen prices we checked, consolidating into 200 costs less.
Regular allowances and reasons to choose each tier
The starting point if one subscription is enough. You do not need a higher tier just to use Sol.
A consolidation option for people with two 100 subscriptions. The regular allowance sits in one account.
For persistent allowance shortages or Astra Ultrafast. A larger regular allowance does not necessarily mean a lower unit cost.
The baseline is current Plus. These multipliers do not guarantee message counts, completed tasks or image counts. They are also distinct from model-specific limits in ordinary Chat.
What this guide covers
What the allowance multipliers tell us
The September 29, 2026 announcement introduced Pro 500 and reopened Pro 200, which had stopped accepting new subscriptions. The current official Pro help page explains the three monthly prices and their relative allowances, but that page alone does not give the exact multipliers. The figures for 100 and 200 are in the representative's original announcement; the figure for 500 is in OpenAI's own announcement.
Plus = 1 → Pro 100 = 5 → New Pro 200 = 10 → Pro 500 = 25
Thus, new 200 is twice 100; 500 is five times 100 and 2.5 times new 200.
These figures describe relative capacity for choosing a Work/Codex allowance. The official pricing documentation says Work and Codex share an allowance, with consumption depending on the model, reasoning effort, input and output, tool use, caching and other factors. Completing 100 tasks on 100 does not guarantee you can complete 200 on 200. Even on the same plan, longer conversations and heavier research change how quickly the allowance is used.
Do not mix the old 200's 20× figure with the new 200's 10×
Search results describing the previous Pro 200 still mention “20 times Plus.” The announcement describes new 200 as 10×, so an old search summary can give you the wrong basis for a subscription decision. The official help page says eligible existing 200 subscribers can keep their previous allowance through October 29, 2026, while maintaining an active Pro 200 subscription, then move to a smaller allowance at the same $200 monthly price.
Eligibility requires an active 200 subscription on the specified cutoff date or within the seven days before it. The current help page does not state that date explicitly, so check your own subscription and the notice you received. Do not assume upgrading from 100 to 200 now gives you the old allowance.
Nor does a change from 20 to 10 prove that the number of tasks you can actually complete falls by exactly half. The representative's original announcement says the baseline allowance is also increasing, without publishing the absolute ratio between old and new Plus. The explanation of the change describes new 200 as roughly half the old 200 in API-spend terms, while also mentioning gains in model efficiency. Cost equivalents, relative multipliers and completed work must be considered separately.
Compare prices and features
The official monthly US-dollar prices are $100, $200 and $500. The Japanese-yen prices below are values the AI Arte editorial team read from one account's Japanese-language web pricing screen using Codex browser controls on September 30, 2026. We only inspected the plan selection display; we did not change a subscription or make a purchase. These are not guaranteed prices for every region or purchase channel, and we did not check the second account's actual bill.
| Comparison | Pro 100 | New Pro 200 | Pro 500 |
|---|---|---|---|
| Official monthly price | $100 | $200 | $500 |
| Tax-inclusive monthly price observed | ¥16,800 | ¥30,000 | ¥84,000 |
| Monthly price × 12, calculated | ¥201,600 | ¥360,000 | ¥1,008,000 |
| Regular allowance versus current Plus | 5× | 10× | 25× |
| GPT-6.1 Sol | Available | Available | Available |
| Standard / Fast | Available on supported models | Available on supported models | Available on supported models |
| Astra Ultrafast | Not included | Not included | Included |
| Work/Codex five-hour limit | None under current Pro guidance | Same | Same |
| Work/Codex weekly limit | May apply; check usage | Same | Same |
| Pro-model allowance in ordinary Chat | GPT-6 Pro and 5.6 Sol Pro share a weekly allowance | Weekly GPT-6 Pro allowance, daily 5.6 Sol Pro allowance and other limits | Specific per-model figures not verified |
| Storage shown in this check | 100 GB | 100 GB | 100 GB |
Sources: Pro tiers, Work/Codex pricing, speed specifications, Pro-model limits in Chat. Twelve-month totals are calculations, not annual subscription offers.
The 100 GB storage figure is file capacity, not monthly tokens or Codex work capacity. Pro generally includes image creation, deep research and other features, but we found no basis for applying the 5/10/25 ratios directly to image counts or ordinary Chat questions. If Chat is your main use, also check the limits and renewal times shown in the model selector.
In ordinary Chat, the two models on 100 share a weekly allowance, whereas 200 has daily and weekly allowances. The explanation says that after the GPT-6 Pro allowance is exhausted, Chat switches to GPT-5.6 Thinking Medium. You can manually select 5.6 Sol Pro if both its model-specific daily allowance and the combined daily allowance for the two models remain available. The current original documentation does not provide a numerical count table, so we have not reused old counts from search results as limits under the new system.
Is one 200 better value than two 100 subscriptions?
If both 100 subscriptions cost ¥16,800 including tax, their combined monthly cost is ¥33,600. The new 200 was shown at ¥30,000 in our check, giving a difference of ¥3,600 per month, or ¥43,200 when multiplied by 12. Their announced regular allowances are also equivalent: 5 + 5 = 10 versus 10. For Codex-focused use with the same model, reasoning effort and speed, consolidating into 200 is the first option to consider.
Equivalent regular allowances. Different yen prices.
Separate balances, histories and payments.
The regular allowance is available in one account.
The difference is ¥3,600 per month, or about 10.7%. This comparison assumes both 100 subscriptions cost the same and excludes temporary benefits.
However, allowances from two subscriptions cannot be pooled into one chat. The official account-switching explanation also says histories, memory and billing remain separate for each account. Consolidating into 200 does not automatically merge another account's previous history. If you need separate work and personal accounts, weigh that convenience alongside the cost difference.
500 is not cheaper per unit of regular capacity than 200
In US dollars, 100/5, 200/10 and 500/25 all equal $20: the same price per announced allowance unit. Using the yen prices observed here, the corresponding values are ¥3,360 for 100, ¥3,000 for 200 and ¥3,360 for 500. Compared with 200, 500 provides 2.5 times the capacity at 2.8 times the price, making its cost per unit of regular capacity 12% higher than 200's by this calculation.
That does not mean 500 has no value. It can be an option if a large allowance in one account or shorter waits with Astra Ultrafast matters to you. Conversely, if Sol at Standard speed is enough, you do not need the most expensive tier simply because it costs the most. See our GPT-6.1 Sol versus GPT-6 Sol comparison for a closer look at model choice and reasoning settings.
How Fast and Ultrafast affect your allowance
Even within one plan, changing the speed setting changes allowance consumption. The official pricing documentation lists different multipliers for the included allowance and purchased credits. Fast is not exclusive to 200 and higher: it is also available on supported models with 100.
| Relative to Standard on the same model = 1 | Included allowance consumption | Credit consumption | What to consider |
|---|---|---|---|
| Standard | 1× | 1× | Baseline for comparing regular capacity |
| Fast | 2.5× | 2× | Available on supported models; use selectively for urgent work |
| Astra Ultrafast | 8× | 6× | Only 500 among Pro tiers; Sol Ultrafast is planned |
Assume the same Astra model and token mix, with all usage on 500 running at Ultrafast. Divide the regular allowance of 25 by the consumption multiplier of 8: 3.125. That is 62.5% of 100's Standard allowance of 5. This is arithmetic using the multipliers, not a measurement of completed tasks.
Ultrafast's “up to 8×” compares token generation speed. It does not guarantee that an entire task, including research, builds and browser operations, finishes eight times faster. The speed specifications checked on September 30, 2026 describe Astra Ultrafast as available, while the DevDay announcement lists GPT-6.1 Sol Ultrafast as forthcoming. We could not establish a launch date.
To test speed, record both time to completion and allowance consumption on a representative task you normally perform. Comparing Astra Ultrafast with Sol Standard mixes model differences with speed differences, so use the same model to isolate the speed setting's effect. One option is Standard for routine work and Fast only when waiting time matters to a deadline.
Free resets, paid resets and credits
Even if you have experienced a temporary allowance reset, do not add it to your expected monthly subscription capacity. Scheduled renewals, distributed free reset coupons, paid immediate resets and extra credits work differently. Which one you use also affects when the allowance next renews.
Use a distributed coupon once before it expires. It renews eligible limits and changes the weekly renewal date. Future distributions are not guaranteed.
Applied immediately upon purchase, even with allowance remaining. Cannot be saved. A new week starts with the next Work/Codex request.
A balance for additional supported usage after the included allowance is used. Does not remove every limit in ordinary Chat.
Official explanations: banked resets / paid weekly resets / credits for personal plans.
A paid reset does not add another week on top of your remaining allowance
Paid reset timeline: an illustrative example
- Purchase on Thursday: The reset applies immediately, even if you have allowance left.
- Next request on Thursday: A new weekly Work/Codex period begins.
- The following Thursday: The next scheduled renewal is seven days after that period starts.
Do not assume you also receive another allowance on the original renewal date.
The official explanation describes a paid reset as bringing forward the regular allowance. Free banked coupons also change the renewal date, so check the new date after using one. A free coupon is consumed only when at least one eligible usage window is successfully reset; it remains available if no window needs resetting. Previous immediate resets granted broadly to users are separate from banked coupons, and are not a guaranteed recurring benefit.
Paid resets are generally non-refundable, and availability can differ by account and billing country even on the same plan. The purchase option in settings is not shown while the eligible allowance is completely unused. A missing option alone does not mean the whole plan is ineligible: check your usage and purchase conditions.
Check your account screen for extra-purchase prices and availability. We did not verify credit or paid-reset purchase prices, so we do not give a yen unit price or claim that buying extra usage is always better value. Buying credits on 100 or 200 does not unlock Ultrafast. If you enable automatic top-ups, also check that additional charges may occur when the balance falls below the configured threshold.
Choose for your workload
Start by using a normal week with the same model, reasoning effort and speed, then identify which tasks leave you short. Do not size your monthly plan solely from a week with a temporary reset or a switch to a new model. If one 100 is enough, it has the lowest fixed cost among the Pro tiers compared here. If you are unsure whether you need Pro at all, consider Plus separately too.
| Your usage | First option to consider | What to check before deciding |
|---|---|---|
| 100 leaves unused allowance every week | One 100 subscription; consider Plus if appropriate | Required features such as Pro models, and normal-week usage |
| Two 100 accounts, mainly for Codex work | Consolidate into new 200 | Both actual bills, benefits, effective date and need for separate histories |
| Allowance runs short only a few days per month | 100 or 200 plus extra usage | Purchase-screen price, remaining allowance and next renewal |
| 200 repeatedly runs short in normal weeks | Compare 500 with 200 plus extra usage | Whether recurring extra costs exceed the ¥54,000 monthly difference |
| You want much shorter Astra generation waits | Ultrafast on 500 | 8× consumption and time saved across the whole task |
| Mostly Pro models in ordinary Chat | Choose after checking Chat limits | Per-model allowances and renewal times, rather than Work/Codex multipliers |
How much extra usage can you buy and still spend less?
Assume both 100 subscriptions cost ¥16,800 and you currently buy no extra usage. After consolidating into 200, you can spend up to ¥3,600 per month on extra usage and keep the total at or below ¥33,600. Spending ¥6,000 extra brings the total to ¥36,000, ¥2,400 above your current cost. With just one 100, extra spending up to ¥16,800 keeps the total no higher than two subscriptions. Cost is only part of the decision: you must also be able to complete the same work when you need to.
New 200: ¥30,000 + that month's extra spending
Current two 100 subscriptions: ¥33,600 + that month's extra spending
500: ¥84,000 + that month's extra spending
Do not treat temporary free benefits as a permanent future contribution; count them separately only in months when they are available.
500 and 200 plus extra usage cost the same when the extra spending reaches ¥54,000 per month. Equal spending does not necessarily buy equal capacity. Compare consumption rates by model and speed, the replacement of remaining allowance by resets, and 500's exclusive features as well.
What to check before changing subscriptions
Look for subscription changes under ChatGPT's plan section in settings, and remaining allowance under usage settings or the Codex usage screen. Labels can differ by device and language. Official help directs you to cancel through the purchase channel you used, so first check whether your subscription is on the web or through Apple or Google.
- Check the actual bill for each subscription. We observed the web price for only one account. If either uses a different channel or price, recalculate the ¥33,600 baseline.
- Match the account whose allowance you are viewing. If the browser and Codex app show different figures, check whether they are signed into the same account. A difference alone does not prove a bug.
- Record free coupons, purchased credits and old 200 benefits separately. If expiry dates or treatment after a plan change are unclear, check the screen or ask support.
- Read the effective date, initial charge and proration on the final purchase screen. We could not establish settlement conditions between the current three tiers. The ¥3,600 monthly difference is not necessarily your first month's saving.
- Confirm that the subscription you no longer need will stop renewing. Removing the app does not cancel it. Cancelling also does not transfer another account's history to the new subscription.
Official guidance says support cannot reset ChatGPT/Codex limits themselves. However, incorrect accounting or being unable to use the service after renewal can be investigated. Do not treat authentication errors or “thread not found” as evidence of an exhausted allowance. To organize model settings, see our guide to changing chat models and reasoning effort.
If you generate many images, also distinguish Codex image-generation allowance consumption from ChatGPT image limits. Our ChatGPT, Codex and Claude image-generation comparison explains the different generation routes and costs.
Conclusion
If you mainly use Codex and have two 100 subscriptions, consolidating into new 200 is a strong option for reducing fixed costs while keeping the announced combined regular allowance. If both 100 subscriptions cost ¥16,800, the 200 price we observed is ¥3,600 less per month. Choose 100 if one subscription is enough; consider 500 if 200 is consistently insufficient or you need Astra Ultrafast.
Your actual yen bills, model, reasoning effort, speed, normal-week usage, extra-purchase prices and temporary benefits can change the decision. Use the published multipliers without inventing unpublished absolute token counts or task counts. Separating the old 200 transition arrangements from new 200 is the starting point for this comparison.
Frequently asked questions
Does Pro 200 provide twice 100's allowance, and 500 five times as much?
That is the relationship between the announced regular allowances, using current Plus as the baseline. It does not guarantee fixed question counts, image counts or completed tasks. Compare the same model, reasoning effort and speed, and check ordinary Chat's model limits separately.
Do I need 200 or 500 to use GPT-6.1 Sol?
No. It is available on 100, and 6.1 Sol is also offered on Plus. Choose a higher tier for the allowance or exclusive features you need. We found no explanation saying the same model and reasoning setting produce better-quality answers solely because the subscription costs more.
Do occasional resets make a cheaper plan better value?
Future distributions are not guaranteed, so do not use them as the basis for recurring costs. A free coupon may reduce spending in a month when you have one, but record it separately from scheduled renewals, paid resets and extra credits.
Did this article measure capacity by switching to 200 or 500?
No. Our independent check was reading one account's Japanese-language web pricing display. The multipliers come from original announcements, and the cost differences are calculations. We did not take out a trial subscription or make paid purchases, and did not measure absolute capacity or completed task counts.
