SSD Nodes Learn Hosting plans →
How to do am Matt ConnorBy Matt Connor

Claude Code cost na how much for Pro, Max and Ultra?

Claude Code na $20/month for Pro, $100 for Max, $200 for Ultra, or per token with API credits. See October 2026 prices and worked estimate.

Claude Code dey cost how much?

Claude Code dey cost $20 every month for Claude Pro plan, $100 every month for Max, $200 every month for Ultra, or you fit pay per token if you use API credits from Claude Console. Free version no dey. Subscription dey buy usage allowance wey Claude Code dey share with Claude chat. API credits dey buy tokens, and dem dey bill am for published rate per million tokens. Your organization monthly spend cap and any lower limit wey you set still apply. Every figure for this page na as of 3 November 2026.

Which option cheap pass depend on how many hours you run am each day, and whether you need report of spending for each user. Figures for both options dey below, together with one worked estimate for one developer wey dey work alone.

Claude Code free dey?

No. Claude Code no get free tier. Anthropic setup documentation talk am directly: Claude Code need Pro, Max, Ultra, Team, Enterprise, or Console account, and the free Claude.ai plan no include Claude Code access. The other option na cloud provider like Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry, wey go bill per token to your cloud account. You fit install the command line tool without paying anything, and claude --version go print version number, but free Claude account no include access to coding sessions. If you already dey pay for Pro and you wan know wetin e cover before you install anything, wetin Pro subscription dey give you for Claude Code answer am plan by plan.

Two things dey make people think say free tier dey. The first na the free Claude.ai plan, wey cover plenty things beyond chat: chat for web, desktop and mobile, web search, creating files and running code, memory across conversations, connecting your apps and tools, Artifacts, and Claude Design, Slides and Docs. Claude Code no dey that list, and Anthropic pricing page mark am as not included on Free. Wetin free Claude plan actually include explain where that boundary dey. The second na the small free credit wey Anthropic give new Claude Console users to test the API. Na starter credit be that, no be monthly allowance, and when e finish, the requests go stop.

The cheapest real entry point na Pro at $20 per month, or $17 per month if you pay for one year upfront.

Subscription price as of October 2026

ChartClaude plan prices in US dollars per month, as of 3 November 2026
Di data wey dey behind dis chart
[
  {
    "plan": "Free",
    "price_usd": 0
  },
  {
    "plan": "Pro",
    "price_usd": 20
  },
  {
    "plan": "Pro, billed annually",
    "price_usd": 17
  },
  {
    "plan": "Max",
    "price_usd": 100
  },
  {
    "plan": "Ultra",
    "price_usd": 200
  },
  {
    "plan": "Team, standard seat",
    "price_usd": 25
  },
  {
    "plan": "Team, premium seat",
    "price_usd": 125
  }
]

Pro wey cost $20 every month na the entry plan wey include Claude Code. Max and Ultra na separate plans, and dem both dey charge monthly only. Max wey cost $100 every month give five times Pro usage every five hours and every week, while Ultra wey cost $200 give ten times the weekly usage. Ultra replace Max 20x, and existing subscribers go move automatically without price change. Na web prices be these; mobile pricing fit different. Team dey charge per person: $25 every month for standard seat and $125 for premium seat. Annual billing cheaper for both. Enterprise dey listed as $20 per seat every month, with annual billing, plus usage charges wey follow API rates. If na group you dey buy for instead of yourself, wetin each Team and Enterprise seat really include compare the seat prices with the cost of paying per token.

Wetin subscription give you na usage allowance, no be token budget wey you fit inspect. Pro and Max get rolling five-hour usage window, plus weekly limit. Ultra get weekly limit, but verified accounts no get five-hour limit; accounts wey never verify get temporary limit. For Max and Ultra, Fable get separate weekly limit too, but that limit dey inside the overall weekly limit, no be extra limit beside am. Anthropic pricing page allow you use up to 50% of the weekly limit for Fable without extra charge. Fable dey use allowance faster than other models, and because e dey draw from the same overall limit as everything else wey you run, you no fit use pass the weekly limit altogether. For Pro, Fable dey use usage credits from the beginning. None of these limits must stop you: when usage credits dey enabled, after you use applicable session limit, weekly limit, or Fable limit for Max or Ultra, your requests switch to pay-as-you-go pricing with usage credits and continue working until you reach your monthly spend limit, if you set one, or your balance finish while auto-reload dey off. Without credits, you go wait for the reset. The plan allowance dey shared with Claude chat and Claude Cowork, so time wey you spend for chat in the morning reduce wetin remain for coding in the afternoon. The complete Claude tier chooser compare the plans with each other in detail.

Wetin API credits cost per token

If you sign Claude Code in with Claude Console account instead, you no go pay monthly fee. You go buy credits, and the tool go spend dem per token. MTok below mean one million tokens.

ChartClaude API token prices in US dollars per million tokens, as of 3 November 2026
Di data wey dey behind dis chart
[
  {
    "label": "Opus 5.5",
    "input_per_mtok_usd": 4,
    "output_per_mtok_usd": 20,
    "cache_read_per_mtok_usd": "0.20"
  },
  {
    "label": "Opus 5",
    "input_per_mtok_usd": 5,
    "output_per_mtok_usd": 25,
    "cache_read_per_mtok_usd": "0.50"
  },
  {
    "label": "Sonnet 5.5",
    "input_per_mtok_usd": 2,
    "output_per_mtok_usd": 10,
    "cache_read_per_mtok_usd": "0.10"
  },
  {
    "label": "Sonnet 5",
    "input_per_mtok_usd": 2,
    "output_per_mtok_usd": 10,
    "cache_read_per_mtok_usd": "0.20"
  },
  {
    "label": "Sonnet 4.6",
    "input_per_mtok_usd": 3,
    "output_per_mtok_usd": 15,
    "cache_read_per_mtok_usd": "0.30"
  },
  {
    "label": "Haiku 4.5",
    "input_per_mtok_usd": 1,
    "output_per_mtok_usd": 5,
    "cache_read_per_mtok_usd": "0.10"
  },
  {
    "label": "Haiku 5.5, prompts up to 100,000 tokens",
    "input_per_mtok_usd": 0.1,
    "output_per_mtok_usd": 0.5,
    "cache_read_per_mtok_usd": "0.01"
  }
]

Opus 5 cost $5 for every million input tokens and $25 for every million output tokens. The newer Opus 5.5 cost less: $4 and $20. Sonnet 5 cost $2 and $10, the same rates as Sonnet 5.5. Anthropic first sell that Sonnet 5 price as introductory rate until 31 August 2026, then make am the standard price and cancel the planned increase to $3 and $15. Haiku 5.5 na the cheapest row now, at $0.10 and $0.50 for prompts up to 100,000 tokens. For longer prompts, the price rise to $0.50 and $2.50. Haiku 4.5 cost $1 and $5.

Max, Ultra and Team plans now include monthly Claude API credits: $100 for Max and $200 for Ultra. But Anthropic no include interactive Claude Code sessions, whether na for terminal, IDE, desktop app or web. E no include extra usage after you reach your plan limits. The credits cover claude -p and Claude Agent SDK when you run dem yourself with API key from the Console organization wey you link to the plan, because Anthropic bill that usage as Agent SDK usage. When claude -p authenticate with your Claude subscription instead, those runs use your plan usage limits and no use the monthly API credits. Runs wey Claude Code GitHub Action, an IDE extension or Claude desktop app start count as Claude Code usage. Dem no dey covered, even with -p.

The headline rate no be the complete bill. Claude Code resend your whole conversation with every request, so most of your input tokens arrive as cache reads. For most models, cache reads cost one-tenth of the base input price: $0.50 per MTok on Opus 5. For Opus 5.5 and Sonnet 5.5, cache reads cost one-twentieth of the base input price: $0.20 per MTok on Opus 5.5. Writing to the cache cost pass base input: 1.25 times the input rate for the five minute cache and 2 times for the one hour cache. Extended thinking dey billed as output tokens, wey be the expensive column above. So if model dey use high effort setting, e go cost real money for every turn.

Model choice na the biggest control wey you get. How to choose which Claude model to run fit change the cost of the same work by factor of five or more.

Which billing path costs less for your usage shape

Subscription fit interactive work. You go sit for terminal, code for some hours, then stop. The bill dey fixed, and the limit wey you go reach na usage window, no be invoice wey you no expect.

API credits fit work wey people dey share or wey you need report per user. Automation by itself no longer force token billing: a GitHub Actions run, including scheduled one, fit authenticate with OAuth token from claude setup-token and use Pro, Max, Ultra, Team or Enterprise subscription, while claude -p still dey use subscription limits when e authenticate with that subscription. But that token dey tied to the subscription of the person wey generate am. So Anthropic recommend API key for secret wey repositories dey share, while per-user cost reporting dey come from the Claude Console dashboard or, for Team and Enterprise, the admin spend report. Both paths fit limit spending. Workspace spend limits for Claude Console fit stop spending when e reach the number wey you choose. For Pro, Max and Ultra, subscription price dey fixed, and you fit set monthly spending cap for any usage credits wey pass the allowance. You fit also set am to unlimited if you no want cap.

The full comparison of API billing against a Claude subscription dey show where the crossover dey, so this page no go repeat the argument. Short answer: if one person dey use Claude Code most working days, subscription dey better. The decision only dey close for light use or use wey come in big bursts. If you still dey choose whether to use a tool, how Claude Code compares with the other coding agents put the subscription prices side by side.

A worked monthly estimate for one developer

Make we take one developer wey dey work alone, about three hours every day, 20 active days for one month, Sonnet as the default model and Opus for the difficult parts. Anthropic publish three figures across enterprise deployments, checked on 3 November 2026: average of about $13 for each developer per active day, $150 to $250 for each developer per month, and cost wey remain below $30 per active day for 90% of users. Multiply the two daily figures by 20 active days to get the monthly amount.

ChartOne solo developer, 20 active days a month: cost by billing path in US dollars, October 2026
Di data wey dey behind dis chart
[
  {
    "label": "Pro subscription",
    "price_usd": 20
  },
  {
    "label": "Max subscription",
    "price_usd": 100
  },
  {
    "label": "Ultra subscription",
    "price_usd": 200
  },
  {
    "label": "API credits, daily average times 20",
    "price_usd": 260
  },
  {
    "label": "API credits, $30 a day times 20",
    "price_usd": 600
  }
]

For token billing, that developer go pay around $260 per month if all the 20 days cost the published daily average. Anthropic own monthly figure lower, $150 to $250 per developer, so $260 dey just above that published range, instead of dey inside am. $600 na wetin 20 days go cost at $30 each, which be the daily figure wey 90% of users stay below. Anthropic no publish am as a monthly percentile, but developer wey days dey regularly pass $30 go pay more than that. Pro at $20 dey far below both figures. Ultra at $200 dey below the daily average when you multiply am out, but e dey inside the published monthly range.

Read the gap carefully, because e no be free money. Pro no dey give you $150 to $260 worth of tokens for $20. E give you an allowance; how quickly you reach the five hour limit depend on the model, conversation length, files and effort level. The honest reading of this table be say Max at $100 per month na the plan wey most likely fit this usage pattern, and e still cost less than the bottom of the published monthly range for token billing. But usage limits still apply, so heavy week fit stop you in the middle of task too. Max and Ultra dey run the same models, but Ultra double the weekly allowance and monthly API credits, and e remove the five hour limit for verified accounts. So choosing between Max and Ultra depend on how often you dey reach session or weekly limits. Every figure here na only half of the total, because wetin plan cost matter less than wetin e save you. whether that monthly price pays for itself at your own hourly rate na the calculation wey worth doing before you choose a tier.

How to check wetin you really dey spend

Run /usage inside one Claude Code session. For subscription, e go show plan usage bars plus breakdown of which skills, subagents, plugins, and MCP servers your recent usage go to. E go also flag any behaviour wey cause 10% or more of that usage. Press d or w to switch between the last 24 hours and the last 7 days.

For API billing, the Session block for top of /usage dey print one dollar figure:

Total cost:            $0.55
Total duration (API):  6m 20s
Total duration (wall): 6h 33m 10s
Total code changes:    0 lines added, 0 lines removed

The system calculate this figure locally from token counts wey use list rates. But if administrator set your contracted rates for managed settings, e go use those rates. By default, e no know about promotional pricing or negotiated discount, so e fit differ from your invoice. The authoritative number na the usage page for Claude Console. The session totals also reset when /clear start new session, so small number after clear no mean say the whole day cheap. How to read Claude Code token usage line by line explain wetin each counter dey count.

When you reach plan limit for Pro, Max, or Ultra, /usage-credits go open usage settings. For there, you fit turn on usage credits and continue work past the allowance with token rates. For Team or Enterprise plan, the same command go open the organization usage settings if you get billing access. If you no get am, e go ask you to confirm, then send request go your organization admins.

Wetín dey push the bill up

Long session dey cost pass the work wey e look like say dey inside am, because every request dey carry the whole conversation. Four habits fit reduce the bill pass anything else:

  • Run /clear between tasks wey no relate. Old context dey send again for every message wey follow am, so if you do unrelated work for one session throughout afternoon, you go pay for the morning work again and again.
  • Switch models with /model. Sonnet fit handle most coding work, and its output rate — meaning price per output token — dey well below Opus own, na Opus be the model wey thinking tokens dey enter.
  • Watch the cache lifetime. If you return to session after long break, e no go hit the cache, so the system go process your full context again and write am back to cache at the cache write rate, wey dey above the base input price. By default, the main conversation cache dey last one hour within subscription included usage, and five minutes when you dey use usage credits or an API key; you fit set either lifetime yourself.
  • Keep agent teams small. Each teammate dey run im own context window, so token usage dey grow roughly in proportion to team size, and team dey use roughly seven times the tokens of standard session when teammates dey run for plan mode.

Keeping the Claude Code context window small na the same skill as keeping the bill small. How much the agent writes matter just as much, because model wey rewrite whole file when three lines for do dey bill every one of those lines at the output rate, and na so e be for skill wey make agent hold to the smallest change wey go work. If you run agents unattended, capping wetin AI agent fit spend for VPS matter even more, because nobody dey watch the session while e dey burn tokens.

VPS dey run am cost extra?

Yes, and e go come as separate bill. If you run Claude Code for server instead of laptop, na the server you go pay for. Claude Code need 4 GB RAM or more, plus supported operating system like Ubuntu 20.04 or later, so small VPS dey enough. Install am the same way you go install am locally:

curl -fsSL https://claude.ai/install.sh | bash
claude --version

If installation work, e go print version string like 2.1.211 (Claude Code). If command report command not found, installer don put binary for ~/.local/bin, but your shell never pick am up: open new shell and try again. Then run claude for your project directory and log in through browser prompt, using the paid account wey you already get. Your plan follow the account, no be the machine, so VPS session dey use the same allowance as your laptop.

The reason to move go server no be cost. Na because server session dey continue even after you close your laptop. Run Claude Code for VPS inside tmux cover that setup, and the permissions to set before you leave agent running cover wetin you need lock down first. Add server cost on top of your plan, and wetin VPS really cost every month give the current range.

FAQ

Claude Code free to use?

No. You need paid Claude plan (Pro, Max, Ultra, Team or Enterprise), Claude Console account wey get API credits, or access through supported cloud provider like Amazon Bedrock, wey dey charge per token. Anthropic setup documentation talk say free Claude.ai plan no include Claude Code access, even though the install itself dey work. New Console users dey receive small free credit to test the API, but e no be recurring free tier.

Claude Pro subscription include Claude Code?

Yes. Pro cost $20 every month, or $17 every month when dem bill am yearly, and e include Claude Code together with Claude chat. The usage allowance dey shared between both, and e reset on rolling five hour window. How fast you reach that window depend on your work, model and effort level. Max at $100 give five times Pro usage every five hours and every week; Ultra at $200 give ten times Pro usage every week and e no get five hour limit for verified accounts.

Subscription or API billing cheaper for daily coding?

For one person wey dey code most working days, subscription almost always cheaper. Anthropic published average across enterprise deployments na about $13 per developer per active day. This na roughly $260 across 20 active days, and $150 to $250 per developer per month, compared with $100 every month for Max. API billing better when you need per user cost reporting or credential for shared automation wey no tie to one person's subscription. Personal subscription no provide either of these. Spend cap no be the main deciding factor, because Pro, Max and Ultra let you set monthly spending cap for usage credits.

How I fit check wetin Claude Code dey cost me?

Run /usage for a session. For subscription, e show how much of your plan allowance you don use and which parts of the session consume am. For API billing, the Session block show dollar total wey e calculate locally with list prices. This fit differ from your real invoice, so use the usage page for Claude Console as the correct amount. The session total reset when you run /clear, so check am before you clear am.

Wetin go happen when I reach my Claude Code usage limit?

Claude Code go tell you which window you reach and when e go reset. Changing models with /model no restore access after session limit or overall weekly limit, because both windows dey shared across models. If you switch away from Fable, you fit continue work with any remaining overall allowance after you reach Fable separate limit. Verified Ultra accounts no get five hour session limit, but their overall weekly limit and separate weekly Fable limit still apply. Run /usage-credits to turn on usage credits and continue after the allowance at token rates. If you no get credits, you go wait until the window reset.

Dis article help you?