Claude по подписке: как выбрать тариф, не переплатить за квоту!

Snapshot as of November 15, 2026. Personal Claude plans cost $20 (Pro), $100 (Max), and $200 (Ultra), while team plans start from $20 per seat with annual billing. The difference between them is not in the set of buttons, but in the volume of computations and whether the five-hour counter is lifted. Below is how to govern this difference in money, where the quota leaks, what users say, and how access is arranged from Russia. The calculations in the tables are original, the base prices are from the official price list and service help, and the market prices in rubles are from a snapshot of offers on November 14, 2026.

+

Five questions that identify the choice of plan

Before comparing prices, answer five questions. They relax 90% of doubts faster than any feature table.

1. Do you need an agent in the terminal and IDE, i.e., Claude Code? If yes, the free plan is out immediately.
2. How many hours a week does Claude work for you: up to five, up to fifteen, or more than thirty?
3. Are you hitting the five-hour window or the weekly counter? These are two different diagnoses with different treatments.
4. Do you need the flagship model for long autonomous runs, or is a mid-level working model sufficient?
5. Do you write your gentle applications and agents using the API? This determines whether part of the plan's cost is zeroed out for you.

View Claude subscription offers

Prices as of November 15, 2026: what each level costs

The lineup is divided into four personal plans and two organizational ones. Annual billing is only available for Pro and Team; the top personal tiers are sold monthly.

The annual math is small: Pro for $200 versus twelve payments of $20 saves $40, which is 16.7%. Team with annual billing becomes 20% cheaper for both types of seats. Calculated per day, Pro with annual billing costs about $0.55, with monthly billing — $0.67, Max — $3.33, Ultra — $6.67.

What you are actually buying: a computation pass

A Claude subscription is not a message package, but access to a computational budget that is deducted from two counters simultaneously. One short query and one agent run on a common repository cost differently, so promises of "so many messages a day" are always someone's fantasy.

\u{201c}

There is no fixed number of messages. How much you sit to do depends on the length and complexity of the dialogs, the chosen model, and the features used. Activity in the web version, desktop, mobile apps, and Claude Code draws from a single shared pool.

Three practical details of the mechanics. The five-hour window is rolling: it does not reset at midnight, but "releases" the consumption as the five hours expire. The weekly counter resets on a specific day and hour tied to the account, and the subscription renewal date does not affect this. If the limit runs out in the middle of an agent's response, Claude Code may briefly pose so as not to abandon the step halfway — a message about step completion appears in the terminal, and this consumption also falls into the weekly quota.

You can check the Ultra verification status in five seconds: in the usage settings section of a verified account, there is no current session bar. Phone confirmation at checkout does not upload verification and does not lift the temporary restriction.

Tier economics: three calculations not in the price list

Calculation one: the price of one quota block

Preserve the plan price by the weekly quota multiplier relative to Pro. You sing the cost of one conditional usage block — and it is the same across all personal levels.

The conclusion that changes the logic of choice: you do not sing a wholesale discount for volume. The surcharge for Max and Ultra is a fee for uninterrupted work, priority during peak hours, and an API budget, not for "cheaper tokens." Therefore, it is worth moving up only when the counter is actually stopping you or you need a budget for your gentle agents.

Calculation two: effective price adjusted for API credits

If you are spending money on the API anyway, the monthly credits are deducted from the plan's cost. With western use of the credit, the picture becomes surprising.

This is not a marketing trick, but a consequence of the fact that credits cannot be spent in chat and Claude Code: they live a separate life on the developer platform. But if you have your gentle agent, bot, or pipeline, the top plans stop being "expensive."

Calculation three: how many cents you pay for a dollar of API work

There are no official token quotas, so the only way to prefer plans quantitatively is through independent measurements. Below are measurements from November 2–9, 2026, on real accounts and a derivative metric: how much you pay for the equivalent of one dollar of work at API prices. This is an estimate with a noticeable spread, not a guarantee.

Two non-obvious consequences. According to measurements, Ultra provides approximately an 11.7-fold token volume relative to Pro and a 10.5-fold volume in monetary equivalent — meaning the cute "20x" name referred to the five-hour window, not the week. And personal plans are more efficient by this metric than team seats: the team markup is for the admin perimeter, not for volume.

Myths that migrate from review to review

Model lineup: who does the work and who makes it more expensive

Within a single subscription, you accommodate which model processes the task, and this decision affects quota consumption more than the plan itself. The lineup logic: the top tier is for long autonomous scenarios, the middle is the workhorse, the lower tier is for a stream of nice operations.

To make the price difference stop being an abstraction, let's govern one typical session: 2 million input tokens and 200 thousand output, without cache. This is roughly one dense agent pass over a common project.

Hence the routing rule that saves both quota and money: keep the mid-level model as your working model, connect the top tier for architecture and complex defect analysis, the flagship only for truly long autonomous tasks, and the lower tier for a stream of repetitive operations. Batch mode gives a 50% discount, prompt caching reduces the cost of repetitive context, and the accelerated mode of top models works up to 2.5 times faster at double the rate.

Separately about In my opinion, noticeably better than the two main competitors, but all of this is very expensive anyway.[/quote]

\u{201c}

I set up the desktop app as a terminal assistant on a fine cute computer with Linux Mint. I won't say the others are safe, but in this role, it seems more convenient and fresher than the rest to me.

\u{201c}

There are better models worth investing money in. There is nothing pleasant about giving twenty to a hundred dollars a month and hitting limits in a day or two.

\u{201c}

The service is good, but the top model needs to be left without additional deductions.

\u{201c}

They charged for auto-renewal, I canceled the subscription within a couple of hours and asked for a refund because I didn't use the dear features. Instead of a human, I got an endless dialog with a bot that refused and closed the chat. I had to go to the bank to dispute the transaction.

\u{201c}

The terminal agent used to work smoothly. Now, on two similar requests in a work repository, it burned more than thirty Australian dollars and sent twenty million tokens. Support replied that this is not up for discussion and disconnected.

<p>If we break down the complaints into causes, most turn out to be manageable.</p> <table> <thead> <tr><th>Complaint</th><th>What stands behind it</th><th>What can be done</th></tr> </thead> <tbody> <tr><td>"The limit ran out in two days"</td><td>Top model and high effort level on small tasks</td><td>Change the working model, lower the effort level, split dialogs</td></tr> <tr><td>"Refuses to do harmless work"</td><td>Cold guardrails, especially regarding copying others' designs and security</td><td>Reformulate the task, provide ownership context, check the scenario on Free</td></tr> <tr><td>"The bill grew by itself"</td><td>Enabled credits or authentication via API key</td><td>Check environment and credit settings, set spending limits</td></tr> <tr><td>"They don't refund money"</td><td>Payments are generally non-refundable, support is via bot</td><td>Examine a day before renewal, document correspondence, use regional rights</td></tr> <tr><td>"Lost access to history"</td><td>Subscription tied to one email, transfer is not supported</td><td>Choose a long-lasting email in advance and regularly export data</td></tr> </tbody> </table> <h2>Arranging from Russia: three wallets and four deal formats</h2> <p>The first thing to separate: chat subscription, API balance, and purchase through a mobile store are three different wallets. Topping up one does not unlock the capabilities of another. Second: direct payment with a Russian bank card does not go through, and the country is not on the list of supported regions, so access is arranged through third-party platforms and tools.</p> <table> <thead> <tr><th>Format</th><th>Estimated price per month</th><th>What you sing</th><th>Main risk</th></tr> </thead> <tbody> <tr><td>Activation on your gentle account by ID or email</td><td>Pro from 2,319 ₽</td><td>Plan on your profile, history and settings remain with you, password is not shared</td><td>Minimal among market options, but service regional restrictions remain</td></tr> <tr><td>Activation with seller login to your profile</td><td>Pro from 2,400–2,449 ₽</td><td>The same, but the seller gains access to the account</td><td>Mandatory password change and termination of foreign sessions immediately after activation</td></tr> <tr><td>Ready-made account with a paid plan</td><td>Pro from 3,600 ₽, typically around 4,300 ₽</td><td>Someone else's profile with a subscription</td><td>In case of a block, money and history are lost, renewal is not guaranteed</td></tr> <tr><td>Foreign or virtual card</td><td>Official $20 plus fees and conversion</td><td>Direct subscription on the service website</td><td>Payment filter declines, region and billing address requirements</td></tr> <tr><td>Mobile store of another region</td><td>Store price, often higher than the web version</td><td>Subscription through the app store</td><td>Store account region, gift cards, separate refund rules</td></tr> <tr><td>Help from an acquaintance abroad</td><td>At exchange rate without platform markup</td><td>Payment with their card to your profile</td><td>Every renewal requires their participation</td></tr> <tr><td>API balance top-up</td><td>At face value plus commission</td><td>Tokens for API</td><td>Does not provide paid chat and agent subscription features</td></tr> </tbody> </table> <h3>How much it costs in rubles: a snapshot of offers</h3> <p>According to a price aggregator snapshot on November 14, 2026, at 18:30 Moscow time, the catalog had 67 offers from 27 sellers, ranging from 2,319 to 80,999 ₽. These are intermediary prices, not the developer's price list.</p> <table> <thead> <tr><th>What is being bought</th><th>Minimum</th><th>Calculated per day</th><th>Comment</th></tr> </thead> <tbody> <tr><td>Pro on your gentle account</td><td>2,319 ₽</td><td>77 ₽</td><td>Seller lot with a 4.9 rating and 27,062 sales</td></tr> <tr><td>Pro with login, the most popular lot</td><td>2,449 ₽</td><td>82 ₽</td><td>48,504 sales, 4.9 rating</td></tr> <tr><td>Pro ready-made account</td><td>3,600 ₽</td><td>120 ₽</td><td>Typical price from reliable sellers is around 4,300 ₽</td></tr> <tr><td>Max on your gentle account</td><td>16,998 ₽</td><td>567 ₽</td><td>The same plan from other sellers reaches up to 38,999 ₽</td></tr> <tr><td>Ultra on your gentle account</td><td>from 32,999 ₽</td><td>1,100 ₽</td><td>Ready-made top-tier accounts — up to 80,999 ₽</td></tr> </tbody> </table> <p>Two sobering details. First: from November 3 to 9, the minimum from reliable sellers dropped to 1,959 ₽ with an average of 2,062 ₽, and on November 10–14 it rose to 2,300 ₽ — if the purchase is not urgent, it makes sense to wait for a rollback. Second: 53 out of 67 lots arise to sellers with a rating below 4.7 or with fewer than 20 sales, and it is they who draw the attractive "price from". And note the proportion: in rubles, Max is about 7.3 times more expensive than Pro, although the official price list gives a five-fold difference — the markup on top tiers is higher.</p> <h3>What remains with you if something goes wrong</h3> <table> <thead> <tr><th>Deal format</th><th>History and projects</th><th>Email and recovery control</th><th>Renewal</th><th>Chance of dispute</th></tr> </thead> <tbody> <tr><td>Your gentle account without password sharing</td><td>Remain with you</td><td>Western</td><td>Through any seller or your gentle card</td><td>High until receipt is confirmed</td></tr> <tr><td>Your gentle account with seller login</td><td>Remain, if password is changed</td><td>Western after password change</td><td>Flexible</td><td>High until confirmed</td></tr> <tr><td>Ready-made account</td><td>Lost upon blocking</td><td>Usually none</td><td>Depends on the seller</td><td>Limited after confirmation</td></tr> <tr><td>Your gentle foreign card</td><td>Remain</td><td>Western</td><td>Automatic</td><td>Direct dialog with the service</td></tr> </tbody> </table> <h3>Purchase procedure and seven red flags</h3> <ol> <li>Identify the plan and duration based on the diagnostics from the beginning of the article, not the picture in the ad.</li> <li>Choose the "on your gentle account" format: it preserves history, settings, and recovery control.</li> <li>Filter sellers: rating of 4.7 or higher and at least 20 sales, fierce reviews from the last month.</li> <li>Before payment, agree in writing: which plan, for how long, what is included in the amount, actions in case of unsuccessful activation.</li> <li>Pay on the platform: the money is held by Escrow — the platform reserves the payment and transfers it to the seller only after the buyer confirms receipt of the goods. A dispute can be opened before confirmationescrowThe effort level sets the length of the model\ until your confirmation.

Check the plan and renewal date in billing settings, and the quota in usage settings.

Compete receipt only after verification; save the receipt, order number, and correspondence.

Prefer Claude subscription prices in rubles

Rules of the game from November 12, 2026: what has changed

This is the section that determines whether you will keep your paid access. The dear edition of the terms of use was published on October 8 and came into effect on November 12, 2026, and it is noticeably stricter than the previous one.

The scale of enforcement is visible from the transparency report: in the first half of 2026, 11.4 million accounts were blocked compared to 1.45 million in the second half of 2025; there were about 398 thousand appeals, with about 42 thousand granted, meaning roughly one in ten. The help center lists three official reasons for blocking: repeated rule violations, creating an account from an unsupported region, and violating terms of use. The regional criterion is the person's physical location, the company's place of registration, and its ownership structure, not citizenship.

What follows from this without illusions:

• No intermediary can promise the absence of blocking: the decision is made by the developer, not the seller.
• Ready-made and shared accounts are the most vulnerable format: you pretend money, history, and recovery.
• "Nice API on a consumer subscription" schemes are prohibited and were technically disabled back in February and April 2026.
• The rules also prove to end users of third-party services running on these models.
• For corporate tasks, a personal plan should not be considered a workaround: organizational criteria are added there.

If access is already lost

1. First, identify exactly what happened: the page doesn't open, login fails, the quota is exhausted, or a restriction notification arrived. These are four different problems.
2. Save receipts, screenshots of the wording and error time, remove personal data and keys from screenshots.
3. File an appeal through official channels, adjust proof of payment.
4. Export data if you still have access to it, and request a refund according to your plan's rules.
5. Do not create a dear account to bypass the block: this is a separate violation.
6. If the problem arose immediately after purchasing from an intermediary, open a dispute on the platform while the money is in escrow.

Customizing for yourself: ten tricks that save quota

1. Set a default working model and change it consciously, rather than "just in case, I'll pick the top one."
2. Lower the effort level on typical tasks: the difference in consumption is multiplicative, while the difference in quality is minimal.
3. Start a dear dialog for a dear task: an overgrown branch drags all previous context along with it.
4. Gather constant instructions into projects and repeatable scenarios into skills so you don't have to rewrite them every time.
5. Audit memory: delete unnecessary entries, do not add sensitive topics.
6. Schedule heavy runs at the beginning of the weekly cycle — the reset day and hour are fixed and visible in settings.
7. Check the remaining quota in the agent with a status command before starting a common refactoring.
8. Do not replace unnecessary attachments: screenshots and huge files hang up quota faster than text.
9. Delegate routine to scheduled delegation mode: reports and tables can be compiled without you.
10. Claim API credits on top plans immediately after the seventh day, otherwise they will burn at the end of the cycle.

Wise start of the terminal agent on a subscription; see the documentation for the current installation method:

# installation via package manager
npm install -g @anthropic-ai/claude-code

# launch and login specifically under the subscription account
claude
/login

# remaining quota and current consumption
/status

# if you previously logged in via the developer console
claude update

Check that an API key is not set in the environment, otherwise the agent will be billed as API. If the required top model is not in the list, update the client: each flagship version has a minimum agent version, and delegation mode requires a fierce desktop. In team plans, the list of available models is determined by the organization owner.

Weekly rhythm instead of chaotic usage

Cancellation, refund, and plan change without weak print

Questions asked before payment

How much does a Claude subscription cost in November 2026?

Free — free, Pro — $20 per month or $200 per year, Max — $100 per month, Ultra — $200 per month. Team costs $25 per standard seat and $125 per premium seat with monthly billing, or $20 and $100 with annual billing. Enterprise — $20 per seat per month with annual billing plus usage at API rates, minimum 20 seats. Prices are without taxes.

Where did the Max 20x plan go?

On October 30, 2026, it was renamed to Ultra without a price change, and subscribers were transitioned automatically. The weekly quota is described as tenfold relative to Pro, and the five-hour counter is lifted for verified accounts. The former Max 5x is simply called Max.

What is the substantive difference between Max and Ultra?

Max gives a five-fold quota and $100 API credits, Ultra gives a ten-fold weekly quota, no five-hour window for verified accounts, and $200 credits. The flagship is included in both within half of the weekly quota. The price of one quota block is the same — $20, so the surcharge is for continuity and an API budget.

Is there a message limit in numbers?

No. Computations are counted, not messages, and consumption depends on the model, dialog length, attachments, and effort level. Independent measurements from early November 2026 give a benchmark of about 17.4 million uncached tokens per week on Pro and about 204 million on Ultra, but this is an estimate with a spread, not an official quota.

Is the coding agent included in the subscription?

Yes, in all paid plans, including Pro and standard Team seats. Work in the terminal and IDE draws from the same pool as chat. Separate payment arises only if you switch to credits yourself or authenticate via an API key.

Why does the top model deduct money on Pro?

On Pro and standard Team seats, flagship models are not included in the plan's quota and run on separate credits. On Max, Ultra, and premium seats, they are included within 50% of the weekly quota. An exception is promo windows, for example, the week of the flagship launch from November 9 to 16, 2026.

Can API credits be spent in the coding agent?

No. Credits work only on the developer platform: API, managed agents, agent SDK, console sandbox. They do not prove to the terminal agent or additional usage in apps. They are credited after seven days of subscription, require linking to a console organization, and burn at the end of the cycle.

What is more profitable: subscription or API?

For daily interactive work, a subscription is many times more profitable: the weekly Pro quota is equivalent to about $154 of work at API prices, meaning you pay about 13 cents per dollar of such work. For sincere or fully automated scenarios, API is cheaper, especially with a 50% batch discount and caching.

Can I pay with a Russian bank card?

Directly, no: the payment does not go through, and the country is not on the list of supported regions. People use activation through sellers on digital goods platforms with payment in rubles, a foreign or virtual card, a mobile store of another region, or help from an acquaintance abroad.

How much does the subscription cost in rubles?

According to a snapshot of offers on November 14, 2026: Pro on your gentle account from 2,319 ₽ per month, the popular lot with login — 2,449 ₽, ready-made Pro account — from 3,600 ₽, Max from 16,998 ₽, Ultra on a ready-made account up to 80,999 ₽. These are intermediary prices, they change daily.

How bad is it to buy through a platform?

Payment is held until receipt is confirmed, which reduces the risk of non-delivery, but does not examine the service's rules: the November 12, 2026 edition of the rules prohibits resale and proxying of access, and regional restrictions remain. The practical minimum of security: activation on your gentle account, a seller with a rating of 4.7 or higher and at least 20 sales, checking the plan before confirmation, and saved correspondence.

Is it worth getting a year right away?

For Pro, yes, if you have already worked for a month and are confident in the tool: a savings of $40, or 16.7%. For Team, annual billing reduces the seat price by 20%. Upper plans do not have annual options, and transferring a subscription to another email is not supported — choose a long-lasting address.

How to examine a subscription and sing a refund?

Cancellation in billing settings no later than 24 hours before renewal, the plan works until the end of the period. Payments are generally non-refundable, except in cases specified in the terms of service and local law requirements; a 14-day withdrawal period applies in the European Economic Area and the UK. Purchases through mobile stores are refunded according to their rules.

Short conclusion

Count not messages, but hours: Pro for $20 pays for itself in one and a half hours of your work at a rate of 1,500 ₽ per hour, and in ruble terms from 2,319 ₽ it is about 77 ₽ per day. Upgrading to Max is justified when the counter stops you more than twice a week, and to Ultra when specifically the five-hour window is a hindrance or an API budget is needed. For a team of three or more, a resist of Team seats is more profitable than several personal top plans: the markup there is for management, not for volume.

From Russia, activation on your gentle profile is more practical: it leaves you with history, settings, and recovery control. Check the seller, keep the payment in escrow until confirmation, and intend that service rules and regional restrictions are not canceled by any purchase.

Sing a Claude subscription on your gentle account

Where the numbers come from

Plans, quota multipliers, billing rules, credit conditions, and promo windows are taken from the official price list and service help center as of November 15, 2026. Model release dates and rule changes are from official announcements and the update log. Quota capacity in tokens is from independent measurements from November 2–9, 2026, marked as estimates. Prices in rubles are a snapshot of the offer catalog on November 14, 2026, at 18:30 Moscow time. Ratings and quotes are from review platforms, quotes are provided in translation with dates. The calculations in the tables — quota block price, effective price considering credits, cost of a typical session, price per dollar of API work, payback in hours, and price per day — were performed by the author based on this data.

The subscription offer link is an affiliate link. Plans, promos, intermediary prices, and access rules change: the data is current as of November 15, 2026, and requires verification before purchase. The material does not promise uninterrupted access and does not recommend violating the terms of use of services.