Cheapest way to use Claude Code: Pro billed yearly at $16.67 a month, as Max and Ultra charge $20 a share

What is the cheapest way to use Claude Code?

Claude Pro billed yearly is the cheapest way to use Claude Code if you code every week. A guest pass is cheaper for the seven days it lasts, and an API key is cheaper if you only code on one or two days a month. We priced each of the routes for one developer, using Anthropic's pricing page, help center and Claude Code docs, plus GitHub's and Cursor's pricing pages, on 9 November 2026. US prices, before tax:

All of the Claude rows share one allowance with Claude chat, and that's the catch most people miss. Anthropic's Pro and Max help page says "Pro and Max plans have a five-hour session limit and a weekly limit," and the chats you have on claude.ai draw on the same two limits. It's also worth buying on the web. Anthropic's help center warns that the prices in its mobile apps can be different.

Do bigger Claude plans make Claude Code cheaper?

No. Max and Ultra charge exactly what Pro charges for each Pro-sized share of usage: $20 a month. Anthropic's Max page gives Max five times Pro's usage, with the multiplier on both the 5-hour and weekly limits. Its Ultra page says "Get 10 times the Pro plan's usage every week" and "The Max 20x plan is now called the Ultra plan." Divide each price by its multiple:

So climbing a tier buys you more room, but it doesn't buy you a lower rate. What Ultra really adds is time. If your Ultra account is verified, it has no 5-hour limit, so on a long day you can draw down the weekly allowance without waiting for a session to reset. Pro billed yearly is the only discount. It's $200 up front, which Anthropic's page rounds to $17 a month, and it takes $3.33 a month off. Max and Ultra are sold monthly only.

There are two cautions. Anthropic doesn't publish how big Pro's weekly allowance is in tokens or in messages, so a "share" is a multiple, not a number that you can count. And the $200 Ultra plan used to be Max 20x, which Anthropic sold as 20 times Pro in each session. Claude Pro vs Claude Max has the signals that say you need the extra room.

Is the Claude API cheaper than Claude Pro for Claude Code?

The API is cheaper than Pro only if you code on about one day a month, at the pace Anthropic says is average. Across enterprise deployments, Anthropic's cost docs say "the average cost is around $13 per developer per active day and $150-250 per developer per month, with costs remaining below $30 per active day for 90% of users." At $13 a day, the API costs more than yearly Pro after 1.3 active days, and more than monthly Pro after 1.5. It passes Max after 7.7 days and Ultra after 15.4.

The model you pick moves that line more than anything else does. The same docs show a sample Claude Code session that used 1,200 input tokens, 5,300 output tokens, 940,000 cache reads and 50,000 cache writes, and it cost $0.55 on Sonnet 4.6. We priced those exact tokens at the rates in Anthropic's pricing docs on 9 November 2026. We counted the cache writes at the 5-minute rate, because that's the rate that makes the $0.55 add up:

Cache reads are the line to watch. On Sonnet 5.5 they make up $0.094 of the $0.27, which is more than the output costs. Claude Code resends your whole conversation with every request, and the docs say that it re-reads the history "at the cached token rate." The Haiku row assumes that each request stays under 100,000 tokens. Above that, Haiku 5.5 bills five times as much. One more caveat: Anthropic's pricing docs say models from Claude 4.7 on use a tokenizer that produces "approximately 30% more tokens for the same text." So the same work on a 5.5 model can cost up to about a third more than these rows, near $0.36 on Sonnet 5.5, which still beats Sonnet 4.6. The month-by-month version of this math, plan by plan, is in Claude Code pricing.

How do you cut Claude Code's token bill?

You cut Claude Code's bill by keeping less in context and sending easy work to cheaper models. These settings come from Anthropic's cost docs, read on 9 November 2026, and they shrink a plan's usage the same way they shrink an API bill:

1. Keep Sonnet 5.5 as the default. Anthropic says Sonnet "handles most coding tasks well and costs less than Opus." Opus 5.5 costs twice as much per token.
2. Run /clear between unrelated tasks. The docs call /clear the fresh start that "costs nothing", while /compact reads the whole conversation it summarizes.
3. Lower /effort for simple jobs. Thinking bills as output, and you can't switch it off on Opus 5.5, Sonnet 5.5 or Haiku 5.5.
4. Put "model: haiku" in subagents that run tests or read logs. Haiku 5.5 costs a twentieth of Sonnet 5.5 per input token.
5. Keep CLAUDE.md under 200 lines and move long instructions into skills, which load only when used.
6. Skip agent teams unless you need them. Anthropic says they use "approximately 7x more tokens than standard sessions" when teammates run in plan mode.
7. Run /status before you blame your plan. If ANTHROPIC_API_KEY is set, Claude Code "will use this API key for authentication instead of your Claude subscription."

Once you're past your plan's limit, it's cheaper to prepay than to pay the list price. Anthropic's usage bundles sell $50 of credit for $45, $250 for $200 and $1,000 for $700, up to $2,000 a month on Pro, Max and Ultra. At 30% off, the Sonnet 5.5 sample session drops to about $0.19. The one balance covers Claude Code, Claude chat and Cowork. The monthly API credits that come with Max won't help you here, though. The help page says that they "can't be used in Claude Code or the Claude apps."

Are Cursor or GitHub Copilot cheaper ways to code with Claude?

GitHub Copilot Pro is cheaper at $10 a month, but what you get is GitHub's agent calling Claude. It isn't Claude Code. GitHub's plans page gives Pro "$15 monthly total credits", at "1 AI credit = $0.01 USD", and lists Claude models from Haiku 5.5 to Opus 5.5. Copilot's third-party agents docs describe a Claude agent that "uses the Claude Agent SDK." GitHub's plans page disagrees with itself on who gets it: the Pro card lists "Access to 3rd party agents (Claude Code and Codex)", while its comparison table includes agent delegation only from Pro+ at $39. Each session "consumes AI credits based on the model used and the number of tokens processed." GitHub doesn't print the rate, so we can't tell you how many sessions $15 will buy.

Cursor costs more for the same Claude work. Cursor's pricing page says third-party models are "not included in the monthly prices shown and must be purchased separately," and bills them at API list price. So Premium is $30 on top of the same Claude tokens that a Console key would bill you for. Cursor Plus at $8 runs Grok models only. Both tools edit code in your editor. Neither gives you Claude Code's terminal agent. Cursor pricing works through a month of Claude in Cursor.

Can you use Claude Code for free?

Only for a week, or if you do it without Claude. A guest pass from someone who has been on Pro or Max for at least 31 days gives 7 days of Pro with Claude Code, if you never had a paid Claude plan or trial. That's the rule in Anthropic's Guest Pass Terms of 5 November 2026. Anthropic's API pricing docs say "New users receive a small amount of free credits to test the API," without naming the amount. Ollama's Claude Code guide runs the Claude Code app on a local open model for $0, but no Claude model answers. Open source maintainers can apply for 6 free months of Ultra. Every route, with who qualifies, is in the Claude Code free trial guide.

Where a chat app can cut a Claude Code bill, and where it can't

A separate chat app helps only when it's your Claude chats that run your plan dry. Claude chat and Claude Code share one allowance, so a long morning of chatting leaves you less for the terminal. Whizi, which publishes this page, is a chat app with no Claude Code, no terminal agent and no API. It can't run Claude Code, and it can't pay for it. What it can do is take the chatting off your Claude plan. Claude Pro billed yearly plus Whizi Pro billed yearly costs $16.67 plus $19.99 a month, and Whizi Pro carries Claude Opus 5.5 and Sonnet 5.5 next to GPT and Gemini. That's less than Max at $100 when chat is why Pro runs out. When code alone burns through Pro, a chat app saves nothing, and Max or the API is the answer. The cheapest way to use Claude and GPT prices the chat side, and one subscription for every AI model shows what a single chat plan covers.