Claude Code Pricing 2026: Pro vs Max and Real Usage Limits

Table of Contents

• How much Claude Code costs
• How the usage limits work
• Pro vs Max: which one to choose
• Subscription or paying per API
• Which model to use with each plan
• How much it costs compared with other AI
• My recommendation

The short verdict: for most people, Claude Code starts with Pro, at 20 USD per month. If you use it every day and the five-hour limit runs out before you finish, the reasonable step up is Max 5x, at 100 USD. Max 20x (200 USD) only makes sense if the agent works with you all day long.

10-second answer: Claude Code has no free plan. It is included in the Claude subscriptions, or you pay per usage with the API.

Plan	Price per month	Who it is for
Pro	20 USD	A few hours of use per week
Max 5x	100 USD	Daily use that hits the Pro limit
Max 20x	200 USD	The agent working all day long
API (Console)	Pay per usage	Sporadic use, automation or companies

Prices verified on October 8, 2026 on the official Claude pricing page and in the official help for the Max plan. Prices do not include taxes and Anthropic can change them.

How much Claude Code costs

Claude Code is included in all paid Claude plans. You do not buy it separately: the subscription is the same one you use in the chat.

Plan	Monthly payment	Annual payment	Claude Code	Usage
Free	0 USD	—	No	Chat only
Pro	20 USD	200 USD per year (17 USD/month)	Yes	At least 5x the Free usage per 5-hour session
Max 5x	100 USD	Not available	Yes	5x the Pro usage per session
Max 20x	200 USD	Not available	Yes	20x the Pro usage per session
Team Standard	25 USD per seat	20 USD per seat per month	Yes	More usage than Pro
Team Premium	125 USD per seat	100 USD per seat per month	Yes	5x the usage of a Standard seat
Enterprise	—	20 USD per seat per month + usage at API rates	Yes	Depending on consumption

Three details worth knowing before you pay:

• Max is paid month to month only. There is no annual discount, unlike Pro.
• Pro includes Opus. You do not need Max to use Opus; what changes is how much of it you can use.
• Enterprise charges the seat plus consumption at API rates, with annual billing.

How the usage limits work

Every plan has limits, and understanding them matters more than the price:

1. Five-hour window. The session limit resets on a rolling five-hour window. It is the one you notice the most day to day.
2. Weekly limit. Paid plans add a weekly cap on top of the session limit. On Max, that weekly cap applies to all models and resets at a fixed time assigned to your account. Max also has a separate weekly limit for Fable.
3. Shared limit. On Pro and Max, the Claude chat and Claude Code draw from the same limits. Everything adds up.

Anthropic does not give a fixed number of messages per plan: consumption depends on the model, the size of the project and how much context the session carries. That is why the plan is described in multiples ("5x", "20x") and not in messages.

Inside Claude Code, the /usage command shows your plan's usage bars and what share of the consumption goes to skills, subagents, plugins or MCP servers. Claude Code warns you when you approach a limit.

What happens when you hit the cap. You have four ways out: wait for the reset, upgrade your plan, activate the extra usage credits of your subscription (with /usage-credits from Claude Code) or buy API credits in the Console. API billing is never activated on its own: you have to accept it.

One practical detail: if the warning is about the session or weekly limit, switching models with /model does not give you access back, because that cap is shared across all models. If the warning is specific to one model ("Opus limit"), you can indeed keep going with another one.

Limits verified on October 8, 2026 in the official help on Claude Code with your Pro or Max plan and in the Claude Code costs documentation.

Pro vs Max: which one to choose

The difference between Pro and Max is not about features: it is about how much you can use Claude Code before hitting the limit.

Stay on Pro if:

• you code with Claude Code a few hours a week;
• you use Sonnet for most tasks and Opus only for the delicate ones;
• you rarely see the five-hour limit warning.

Move to Max 5x if:

• the five-hour limit runs out almost every day mid-task;
• you work with large projects, where every message loads a lot of context;
• you use Opus as your main model.

Move to Max 20x if:

• the agent works with you all day long;
• you run several sessions or subagents in parallel;
• Max 5x also falls short before the week is over.

A simple rule: start on Pro for two weeks and watch /usage. If you hit the limit several days a week, step up to Max 5x. It is cheaper to find out this way than to pay for Max just in case.

Subscription or paying per API

With an Anthropic Console account, Claude Code charges for the tokens consumed, at API rates:

Model	Input (USD per million tokens)	Output (USD per million tokens)
Fable 5.1	10	50
Opus 5.5	4	20
Sonnet 5.5	2	10

For reference, Anthropic reports that in companies the average cost with API is around 13 USD per developer per active day and between 150 and 250 USD per developer per month, and that 90% of users stay below 30 USD per active day.

With those numbers, if you use Claude Code daily, Pro or Max 5x usually works out cheaper than the API. The API makes sense in three cases:

• your usage is sporadic and does not justify a fixed fee;
• you automate it in CI or in scripts, with no person at the keyboard;
• your company already pays for consumption, directly or through Amazon Bedrock, Google Cloud or Microsoft Foundry.

API rates verified on October 8, 2026 on the official Claude pricing page.

Which model to use with each plan

With Pro, Max, Team, Enterprise or the API, Claude Code uses Opus 5.5 by default. You can change it with /model.

The model directly affects how long your limit lasts. Sonnet handles most coding tasks well and consumes less than Opus. Opus is best saved for architecture decisions or multi-step reasoning. If you fall short on Pro, try using Sonnet as your main model first before paying for Max.

I compare the models with prices and use cases in Opus 5.5 vs Sonnet 5.5: which Claude model to use.

How much it costs compared with other AI

If you want to see Claude Code in context, the comparison of subscriptions, API and local models across Claude, ChatGPT, Gemini and open models is in real AI prices in 2026.

If you are looking for another subscription to code in the terminal, with models from other providers, check out OpenCode Go price and limits.

If you're torn between OpenCode and Claude Code, I compare price, models and limits in OpenCode vs Claude Code.

My recommendation

Start with Pro. Use Sonnet for the day to day and save Opus for the hard stuff. Check /usage every few days during the first two weeks. If the limit slows you down almost daily, step up to Max 5x; if it still falls short, Max 20x. Leave the API for automations or for when your company pays for consumption.

If you do not have it installed yet, start with the Claude Code guide.