Back to guides
Prices·October 3, 2026·Updated October 4, 2026·12 min read

Claude Code price 2026: Pro and Max subscription or API key, session cost with caching, and payment with BLIK

Claude Code can be paid for with a Claude subscription or an API key charged by token. Here you will find the official plan prices, the cost of an example session with caching, the point at which each option becomes worthwhile, and payment by BLIK for the key.

Claude Code itself costs nothing—you pay for the model powering it, and you have two ways to do that. With a subscription, Claude Code is included in the plan: Claude Pro costs $20 per month ($17 per month with annual billing, paid upfront at $200), while Claude Max costs $100 or $200 per month. These are Anthropic's US prices before tax; Claude's free plan does not include Claude Code, and a subscription purchased through the website can be paid for only by card. With an API key, there is no subscription, only a per-token price: in Kunavo, Claude Sonnet 5 costs $1,40 per million input tokens and $7,00 per million output tokens, so a typical agent step (about 25 000 input tokens and 1200 output tokens) costs approximately $0,0434. You can top up your Kunavo balance with BLIK starting from $10.

Which option is cheaper comes down to one division: $20 for Pro equals approximately 460 such steps per month in Kunavo. If you work with Claude Code almost every day and Pro's limits are sufficient, Pro will be cheaper. If you work in bursts, the key will cost less—a month without coding costs $0.

As of 3 października 2026. Ceny planów pochodzą z claude.com/pricing (strona nie ma wersji polskiej) oraz z centrum pomocy Anthropic: Pro plan, Max plan i Payment FAQ. Ceny Kunavo pochodzą z katalogu modeli (stan na 4 października 2026). Kwoty podajemy w dolarach, bo w dolarach publikują je Anthropic i Kunavo.

How much does Claude Code cost? Subscription or API key

Subscription (Pro / Max)API key (per token)
What you payFixed monthly fee from $20 (US price, before tax)Per token, with no monthly fee
LimitsA session limit renewed every five hours plus a weekly limit, shared with Claude chatNo limit windows; your balance is the boundary
Light monthFull fee$0
Intensive monthWait for the reset or use usage credits at API ratesWork continues, and the cost grows with usage
How you payThrough the website, by card only; in the app, through the App Store or Google PlayIn Kunavo: BLIK, card, Link, Apple Pay, Google Pay
Cost visibilityUsage bars in /usageTo the token, for each key
Who it's forConsistent daily work by one personBurst-based work, per-person budgets, CI, and automation

Obu sposobów nie da się połączyć w jednym rozliczeniu. Gdy ustawisz zmienną ANTHROPIC_AUTH_TOKEN (albo ANTHROPIC_API_KEY, w trybie interaktywnym po jednorazowym zatwierdzeniu), Claude Code używa klucza zamiast subskrypcji: limity planu przestają obowiązywać, a za tokeny płaci właściciel klucza (LLM gateway documentation i article about Claude Code in Pro and Max plans, sprawdzone 3 października 2026). Zapisane logowanie claude.ai zostaje, ale nie jest używane; po usunięciu zmiennej Claude Code wraca do subskrypcji (gateway connection documentation). Który sposób jest aktywny, pokazuje polecenie /status.

Claude Code with a subscription: Pro, Max, and Team

Według installation documentation Claude Code wymaga konta Pro, Max, Team, Enterprise albo Console; darmowy plan claude.ai go nie obejmuje. Polska jest na list of countries, w których Anthropic udostępnia zarówno API, jak i Claude.ai. Ceny według claude.com/pricing i artykułu o planie Max, w dolarach, bez podatku (sprawdzone 3 października 2026):

PlanPrice (USD, before tax)Usage scope
Free$0Without Claude Code
Pro$20 per month; $17 per month with annual billing ($200 upfront)Basic limit, Claude Code included
Max 5x$100 per monthFive times more usage per session than Pro
Max 20x$200 per monthTwenty times more usage per session than Pro
Team, standard seat$20 per seat per month with annual billingPlan for teams

According to Anthropic, Max prices apply to subscriptions purchased through the website; they may differ in mobile apps. According to claude.com/pricing, Fable models are available in Pro only through usage credits (paid usage at API rates), and in Max they can consume up to 50% of the weekly limit.

How much is that in złoty, and what about VAT?

Anthropic podaje ceny w dolarach i zaznacza na stronie z cenami, że nie zawierają podatku. Według article about the Pro plan ceny miesięczne różnią się w zależności od regionu, a w części regionów podatek jest już w wyświetlanej cenie, w innych doliczany przy zakupie. Podstawowa stawka VAT w Polsce wynosi 23% (Your Europe, European Commission, sprawdzone 3 października 2026). Według Anthropic jurysdykcję podatkową wyznacza adres rozliczeniowy (article about the VAT number, sprawdzony 3 października 2026).

Amounts in złoty circulating in Polish guides differ because everyone calculates them differently: dollars multiplied by the exchange rate plus VAT, the euro price, or the App Store price. Purchases in the iOS or Android app are billed by Apple or Google, so they may cost differently than purchases through the website. The binding amount is what Anthropic shows at checkout—which is why we do not provide a złoty price here.

You cannot pay for Pro or Max with BLIK. Według Anthropic payment FAQ subskrypcje kupione przez stronę można opłacić tylko kartą kredytową lub debetową, a przy zakupie w aplikacji płatność obsługuje App Store albo Google Play. Numer VAT można, zależnie od kraju, podać przy zakupie Pro lub Max albo później w Settings > Billing (artykuł o numerze VAT).

Limits: what you are really paying for in a subscription

Pro and Max measure usage with a session limit that renews every five hours and a weekly limit for all models (article about the Pro plan). Both limits are shared by Claude chat and Claude Code: everything you do in either one counts toward the same pool (article about Claude Code in Pro and Max plans). Anthropic does not publish specific numbers; the Claude limits guide describes how the limit windows work in detail (in English).

Gdy limit się skończy, możesz poczekać na reset, przejść na wyższy plan albo włączyć usage credits. Są one przedpłacone i rozliczane po standardowych stawkach API (Anthropic help center), a cache żyje wtedy domyślnie pięć minut zamiast godziny (cost documentation, obie sprawdzone 3 października 2026). Once you exceed the limit, you pay the API price anyway – i tu przydaje się własny klucz: w Kunavo modele Claude są o 30% tańsze niż w cenniku Anthropic.

Is Claude Code free?

The tool, yes; using it, no. CLI i rozszerzenia do edytorów nic nie kosztują; płacisz za wywołania modelu, również te, które wykonuje subagent. Darmowej drogi nie ma: plan Free nie obejmuje Claude Code, a API rozlicza tokeny od pierwszego zapytania. Kredyty kupione bezpośrednio w API Anthropic są przedpłacone, wygasają rok po zakupie i nie podlegają zwrotowi (Anthropic help center, sprawdzone 3 października 2026).

The least binding option is a prepaid key: on Kunavo, top-ups start at $10, there is no monthly fee, and the balance never expires. For $10, on Claude Sonnet 5 you can complete about 230 agent steps, or approximately 12 tasks of 20 steps (not counting cache).

Claude Code through the API: per-token pricing

Z kluczem nie ma planu ani okien limitów: Claude Code jest rozliczany jak każdy inny ruch API, za tokeny. Pliki i wyniki narzędzi liczą się jako wejście, a edycje, wyjaśnienia i tokeny myślenia (thinking) jako wyjście. Ceny w dolarach za milion tokenów (wejście / wyjście); cennik Anthropic z claude.com/pricing (sprawdzony 3 października 2026), ceny Kunavo z katalogu (sprawdzone 4 października 2026):

ModelAnthropic pricingKunavoLess byOne stepIn Claude Code, up to…
Claude Haiku 4.5
claude-haiku-4-5
$1,00 / $5,00$0,70 / $3,5030%$0,0217Background tasks, summaries, simple subagents
Claude Sonnet 4.6
claude-sonnet-4-6
$3,00 / $15,00$2,10 / $10,5030%$0,0651Previous generation, for existing configurations
Claude Sonnet 5
claude-sonnet-5
$2,00 / $10,00$1,40 / $7,0030%$0,0434Daily work (the default model in the configuration below)
Claude Opus 5
claude-opus-5
$5,00 / $25,00$3,50 / $17,5030%$0,1085Previous Opus, for existing configurations
Claude Opus 5.5
claude-opus-5-5
$4,00 / $20,00$2,80 / $14,0030%$0,0868Planning and changes across multiple files (opus alias below); thinking cannot be disabled

One step is one exchange with the model within a task: the system prompt, history, and fresh context from files on input, and an edit or explanation on output—approximately 25 000 input tokens and 1200 output tokens, excluding cache. All language versions of this guide use the same measure.

Cache determines the bill

Claude Code sends the entire conversation with every step. Most of the input is therefore repeated context, and it becomes cheaper through prompt caching. This is how Kunavo charges for cache (per million tokens, with a 5-minute window; the multiplier refers to the input price of the given model):

ModelCache readCache writeInput without cache
Claude Haiku 4.50,1× = $0,071,25× = $0,875$0,70
Claude Sonnet 4.60,1× = $0,211,25× = $2,625$2,10
Claude Sonnet 50,1× = $0,141,25× = $1,75$1,40
Claude Opus 50,1× = $0,351,25× = $4,375$3,50
Claude Opus 5.50,05× = $0,141,25× = $3,50$2,80

A coffee break costs money. Z kluczem API cache żyje domyślnie pięć minut, w subskrypcji godzinę (cost documentation, sprawdzona 3 października 2026). Po dłuższej przerwie pierwsza wiadomość przetwarza cały kontekst od nowa, jako zapis do cache zamiast odczytu. Przy 200 000 tokenów kontekstu na Claude Sonnet 5 samo ponowne wczytanie kosztuje w Kunavo około $0,35 zamiast $0,028. Jeśli po przerwie i tak bierzesz się za coś nowego, wpisz najpierw /clear. Jak działa cache, wyjaśnia cache documentation (po angielsku).

Example session: how much it costs with cache

Poniżej przeliczenie, które możesz powtórzyć. Punktem wyjścia jest przykładowy wynik polecenia /usage z Anthropic's cost documentation (sprawdzonej 3 października 2026). Założenia:

  • The token counts come directly from the documentation: 1200 input, 5300 output, 940 000 cache reads, and 50 000 cache writes. At Anthropic, this is a Sonnet 4.6 session priced at $0.55 according to the pricing page.
  • We recalculate the same figures at the rates of four models; only the price changes. We omit tokenizer differences between models.
  • We calculate cache writes at the rate for a 5-minute window because that is how long the cache lives by default with an API key.
  • Thinking tokens are billed as output (according to the same documentation), so more thinking increases the output portion.
  • With no background calls; according to the documentation, this is usually below $0.04 per session.
ModelAnthropic pricingKunavoCache-read share of the Kunavo amount
Claude Sonnet 5$0,368$0,258approx. 51%
Claude Opus 5.5$0,549$0,384approx. 34%
Claude Opus 5$0,921$0,645approx. 51%
Claude Haiku 4.5$0,184$0,129approx. 51%

On Claude Sonnet 5, this session costs about $0,258 on Kunavo, compared with about $0,368 on Anthropic's pricing; cache reads are approximately 51% of the Kunavo amount. $20, the Pro price, is enough for about 77 such sessions on Kunavo. Claude Opus 5.5 costs about $0,384 in the same example, Claude Opus 5 about $0,645, and Claude Haiku 4.5 about $0,129.

This is example arithmetic, not a measured bill or an upper cost limit; your sessions may be much longer or shorter. /usage in Claude Code shows the four token counts for the current session—insert them into this script:

koszt_claude_code.py
# Koszt sesji Claude Code z jej własnych liczb tokenów (USD, ceny Kunavo).
# /usage w Claude Code pokazuje cztery liczby bieżącej sesji.
STAWKI = {  # za 1M tokenów: (wejście, wyjście, odczyt z cache, zapis do cache)
    "claude-sonnet-5": (1.40, 7.00, 0.14, 1.75),
    "claude-opus-5-5": (2.80, 14.00, 0.14, 3.50),
    "claude-opus-5": (3.50, 17.50, 0.35, 4.375),
    "claude-haiku-4-5": (0.70, 3.50, 0.07, 0.875),
}

def koszt_sesji(model, wejscie, wyjscie, odczyt_cache, zapis_cache):
    i, o, r, w = STAWKI[model]
    return (wejscie * i + wyjscie * o + odczyt_cache * r + zapis_cache * w) / 1e6

# Przykładowa sesja z dokumentacji kosztów Anthropic:
print(round(koszt_sesji("claude-sonnet-5", 1_200, 5_300, 940_000, 50_000), 3))  # 0.258

The dollar amount shown by /usage is an estimate calculated locally by Claude Code using Anthropic's pricing; with Kunavo, the amount actually charged is visible in the usage history in the dashboard. You can calculate your own scenarios in the Claude token cost calculator.

For scale: according to the cost documentation, the average cost in enterprise deployments is approximately $13 per developer per day of active work and $150–250 per developer per month, and nine out of ten users do not exceed $30 per day of active work. These are averages from enterprise deployments, not Kunavo prices.

Subscription or API: which should you choose

Both options on one scale: monthly fee ÷ cost per step = number of steps per month at which the subscription pays off. Using Anthropic's prices before tax and excluding cache:

PlanFee (before tax)Steps per Claude Sonnet 5 in Kunavo ($0,0434 per step)
Pro$20 per month460 steps (approximately 23 tasks of 20 steps)
Max 5x$100 per month2300 steps (approximately 115 tasks of 20 steps)
Max 20x$200 per month4610 steps (approximately 230 tasks of 20 steps)

With warm cache—when 20 000 of the 25 000 input tokens come from cache and the rest are written—a step costs approximately $0,020, so $20 gets you approximately 1000 steps. An intensive month (20 days at 100 steps) on Claude Sonnet 5 costs approximately $87 in Kunavo without cache and approximately $40 with warm cache. Anthropic does not publish how many steps fit within a plan's limits; this comparison only shows the number of steps from which the flat fee is cheaper—provided the limits allow that much.

  • Pro pays off quickly. With a $20 key, you can make approximately 460 steps, or roughly 5 dni of intensive work. Anyone who works almost every day and stays within Pro's limits will pay less with Pro.
  • With Max, the threshold is much higher. Max 5x pays off on Claude Sonnet 5 only from approximately 2300 steps per month, or roughly 23 dni of intensive work.
  • Anyone who works in bursts stays below these thresholds and pays less with a key, because quiet weeks cost nothing. A key also suits CI, automation, and per-person team budgets.
  • After reaching the limit, you can stay on Pro and set a key for the blocked hours instead of buying usage credits at API rates.

Does Max pay off?

Max 5x costs $80 more than Pro. For that amount, you can get approximately 1840 steps on Claude Sonnet 5 in Kunavo, excluding cache—about 37 zablokowanych popołudni of 50 steps, each costing approximately $2,17. Max 5x therefore pays off only when the work blocked by Pro's limits would cost more with a key. If you reach the limit less often, it is cheaper to stay on Pro and set a key only for those hours.

Claude Code configuration with Kunavo

Claude Code reads ANTHROPIC_BASE_URL, so it connects to Kunavo through the Anthropic Messages API without additional plugins. For one terminal session or permanently in ~/.zshrc:

~/.zshrc
export ANTHROPIC_BASE_URL=https://api.kunavo.com   # sam adres, bez /v1
export ANTHROPIC_AUTH_TOKEN=sk-kn-...
export ANTHROPIC_MODEL=claude-sonnet-5
export ANTHROPIC_DEFAULT_OPUS_MODEL=claude-opus-5-5
export ANTHROPIC_DEFAULT_SONNET_MODEL=claude-sonnet-5
export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4-5

ANTHROPIC_BASE_URL to sam adres serwera; ścieżkę /v1/messages Claude Code dopisuje sam. Użyj ANTHROPIC_AUTH_TOKEN, a nie ANTHROPIC_API_KEY: pierwsza zmienna trafia do nagłówka Authorization: Bearer, druga do x-api-key (gateway connection documentation, sprawdzona 3 października 2026). Kunavo przyjmuje klucz z obu nagłówków, ale ANTHROPIC_AUTH_TOKEN działa od razu, a ANTHROPIC_API_KEY w trybie interaktywnym trzeba raz zatwierdzić; odrzucony wcześniej klucz Claude Code potem po cichu pomija. Docelowo zmienne warto wpisać do bloku env w ustawieniach użytkownika; gdy ta sama zmienna jest i w powłoce, i w pliku ustawień, wygrywa plik:

~/.claude/settings.json
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.kunavo.com",
    "ANTHROPIC_AUTH_TOKEN": "sk-kn-...",
    "ANTHROPIC_MODEL": "claude-sonnet-5",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-5-5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-haiku-4-5"
  }
}

Never enter the key in the .claude/settings.json project file—this file goes into the repository. In the VS Code extension, set variables in claudeCode.environmentVariables in the VS Code user settings.

Why we pin models. Według model configuration documentation (sprawdzonej 3 października 2026) Claude Code na API Anthropic domyślnie startuje na Opus 5.5, alias opus wskazuje Opus 5.5, a alias sonnet – Sonnet 5.5; aliasy z czasem przechodzą na nowsze wersje. Sonnet 5.5 is not available in Kunavo, so without ANTHROPIC_DEFAULT_SONNET_MODEL, the /model sonnet command, the execution stage in opusplan mode, and subagents with model: sonnet configured would send requests to a model Kunavo does not support and would end with a 404 error. ANTHROPIC_MODEL ustawia tańszy Claude Sonnet 5 jako model startowy, a ANTHROPIC_DEFAULT_HAIKU_MODEL decyduje też o modelu zadań w tle. ANTHROPIC_DEFAULT_OPUS_MODEL kieruje alias opus (i etap planowania w opusplan) na Claude Opus 5.5 ($2,80 / $14,00). Ten model wymaga Claude Code w wersji 2.1.280 lub nowszej; starszą wersję zaktualizujesz poleceniem claude update. Na Claude Opus 5.5 nie da się wyłączyć myślenia, a tokeny myślenia są rozliczane jako wyjście (cost documentation).

Run claude and enter /status: the Anthropic base URL row should show https://api.kunavo.com, and the Auth token row should show the ANTHROPIC_AUTH_TOKEN variable. If you instead see Login method with a claude.ai account, the variable did not reach the session. Full instructions (in English): Claude Code with Kunavo. Create a key beginning with sk-kn- after signing up in the key dashboard.

Payment by BLIK and card—API key only

Kunavo to bramka API zgodna z interfejsami Anthropic i OpenAI. Doładowanie salda przechodzi przez stronę płatności Stripe. Ceny są w dolarach, ale Stripe pokazuje kwotę w złotych, zanim zapłacisz. Według Stripe Adaptive Pricing documentation kurs zawiera 2–4% opłaty za przewalutowanie, którą płaci kupujący; nie płacisz jej, jeśli wybierzesz płatność w dolarach – wtedy kurs i opłaty może naliczyć Twój bank (sprawdzone 3 października 2026). Przy płatności w złotych do wyboru są:

  • BLIK – wpisujesz sześciocyfrowy kod z aplikacji bankowej, ważny 2 minuty (Stripe BLIK documentation). BLIK działa tylko przy płatności w złotych.
  • Cards (Visa, Mastercard) and Link.
  • Apple Pay in Safari and on iOS, Google Pay in Chrome and on Android, if you have Google Pay configured.

The following are unavailable: Przelewy24, PayU, PayPal, ordinary bank transfer, and SEPA direct debit. Kunavo does not issue VAT invoices; top-up history (date, amount, status) is available in the Billing dashboard.

  • Minimum top-up $10; larger amounts receive a bonus: $100 → $110 na saldzie, $1000 → $1200 na saldzie, $5000 → $6250 na saldzie.
  • The balance is prepaid, has no monthly fee, and does not expire; failed requests are not charged.
  • Automatic top-up works only with a saved card or Link; with BLIK, you top up manually each time.
  • These methods top up your Kunavo API balance. You cannot use them to pay for Claude Pro or Max: when purchasing through the Anthropic website, only cards are accepted.

Step by step: how to pay for Claude with BLIK. Ready to start? Create a Kunavo account, top up your balance, and create a key. Prices for all models are in the Kunavo pricing.

What Kunavo does not replace

Rdzeń Claude Code – agent, narzędzia, subagenty, serwery MCP – korzysta z Messages API, więc działa też z kluczem bramki. Remote Control i dyktowanie głosowe wymagają jednak tożsamości claude.ai i z kluczem bramki nie działają (gateway connection documentation, sprawdzona 3 października 2026).

Kunavo obsługuje Claude na współdzielonej przepustowości, bez dedykowanej puli i bez umownego SLA; kto potrzebuje gwarantowanej przepustowości, powinien kupić ją bezpośrednio u Anthropic. Anthropic nie popiera, nie utrzymuje i nie audytuje bramek innych firm (LLM gateway documentation). I jeszcze raz: kto koduje codziennie i mieści się w limitach Pro, prawdopodobnie zapłaci mniej za Pro. Klucz opłaca się przy pracy zrywami, w zespołach z budżetem na osobę, w CI i na godziny po wyczerpaniu limitu.

Frequently asked questions

How much does Claude Code cost?

Claude Code itself is free; you pay for the model, either through a subscription or by token. With a subscription, Claude Code is included in Claude Pro ($20 per month or $17 per month with annual billing) and Claude Max ($100 or $200 per month), as well as Team and Enterprise plans. These are Anthropic’s US prices excluding tax according to claude.com/pricing (as of 3 października 2026); prices and currency may differ in other regions. With an API key, you pay for tokens without a subscription: at Kunavo, Claude Sonnet 5 costs $1,40 / $7,00 per million input and output tokens, so an agent step (approximately 25 000 input and 1200 output tokens) costs about $0,0434.

Is Claude Code free?

Only the tool itself. The CLI, editor extensions, and configuration cost nothing; model calls do. According to Anthropic’s documentation, the free claude.ai plan does not include Claude Code, and the API charges for tokens from the first request. Without a subscription, you can use a prepaid key: at Kunavo, top-ups start at $10, the balance does not expire, and a month without coding costs $0.

How much does Claude Pro cost in Polish zloty, and does the price include VAT?

Anthropic lists the Pro price in dollars: $20 per month in the US, excluding tax. According to Anthropic’s help center, monthly prices vary by region; in some regions, tax is already included in the displayed price, while in others it is added at purchase. The standard VAT rate in Poland is 23% (European Commission, Your Europe), and according to Anthropic, the tax jurisdiction is determined by the billing address. Zloty amounts in Polish guides differ because each calculates them differently; the amount shown at purchase is binding. Purchases in mobile apps are billed by Apple or Google, so the price may differ from the website.

Can I pay for Claude Code with BLIK?

Not for a Claude Pro or Max subscription: according to Anthropic’s payment FAQ, subscriptions purchased through the website can only be paid for with a credit or debit card, while in-app purchases are handled by the App Store or Google Play. You can pay with BLIK through the API route: topping up your Kunavo balance goes through Stripe’s payment page, which shows BLIK, cards (Visa, Mastercard), Link, Apple Pay, and Google Pay when the amount is in zloty. Stripe’s exchange rate includes a 2–4% currency conversion fee. Przelewy24, PayU, PayPal, and a regular bank transfer are unavailable, and Kunavo does not issue VAT invoices. The minimum top-up is $10.

Subscription or API key—which is more cost-effective?

Divide the monthly subscription fee by the cost of one agent step. On Kunavo, a step with Claude Sonnet 5 costs about $0,0434, so $20 for Pro equals about 460 steps per month (approximately 23 tasks of 20 steps), while $100 for Max 5x equals about 2300 steps. If you work almost every day and stay within Pro's limits, Pro will cost less. If you work in bursts, stay below these thresholds, and pay less with a key, because quiet weeks cost nothing.

Is the API included in Claude Pro?

No. The Claude subscription and Claude API are separate products with separate billing. When ANTHROPIC_AUTH_TOKEN or ANTHROPIC_API_KEY is set in Claude Code, Claude Code uses the key instead of the subscription: plan limits do not apply, and the key owner pays for the tokens. Saved claude.ai login remains; after removing the variable, Claude Code returns to the subscription. The /status command shows which method is active.

How much does a single Claude Code session through the API cost?

It depends on the number of tokens. An example session from Anthropic's cost documentation (1200 input tokens, 5300 output tokens, 940 000 cache read tokens, and 50 000 cache write tokens, recorded on Sonnet 4.6) costs approximately $0,258 in Kunavo at Claude Sonnet 5 rates and approximately $0,368 at Anthropic's listed prices; cache reads account for roughly 51% of the Kunavo amount. This is a calculation based on the example, not a measured bill: differences in tokenizers are ignored. The /usage command in Claude Code shows your own figures.

What should you do when you reach the limit in Claude Code?

With a subscription, work stops until the limit resets: the session limit renews every five hours, and the weekly limit renews once a week. According to the Claude Code cost documentation, both apply to all models, so switching models does not help with them; it only helps with a limit for a single model family (for example, an Opus limit message). Three options remain: wait, switch to a higher plan, or continue working for a fee by enabling usage credits (prepaid, at standard API rates) or setting your own API key, which Claude Code uses as long as the variable is set.