Claude Plans 2026: Pro vs Max, Usage Limits, ChatGPT Prices

Last updated: 12 September 2026. Prices and limits below are from Anthropic's and OpenAI's own pages, checked that day.

A Claude plan is a monthly subscription that sets how much you can use Claude, Claude Code and the other Claude apps before you have to wait for a limit to reset. The price tells you the tier; the limits tell you how much work fits in it, and that second part is what most people are actually searching for.

I have been on Claude Max 20x since January 2026 and use it almost entirely through Claude Code. For most of that time I worked on Sonnet, and it was enough. When Opus became the default model, my token use jumped, and I now switch back to Sonnet for routine work to make the allowance last.

If you want Claude through the API instead of a subscription, see Claude API pricing and Claude free credits. For Claude Code specifically, see Claude Code pricing.

Claude plans and prices (official, September 2026)

Every personal plan has a 200K-token context window and includes Sonnet and Haiku, web search, memory, artifacts, projects on paid plans, and file creation with code execution. Max adds higher output limits and priority access at high-traffic times. Sources: Anthropic's pricing page and the Help Center article What is the Max plan?, both checked 12 September 2026.

Three details that are easy to miss:

• Max is monthly only. Pro can be billed annually; Max cannot. If you move from annual Pro to Max and your remaining Pro balance is larger than the Max price, Anthropic credits the difference to future bills.
• Prices are for web subscriptions. Anthropic notes that mobile pricing may vary by app store.
• There is no standing discount on Max. Anthropic says it occasionally runs limited-time promotions but has no regular discounted pricing on any paid plan.

Team and Enterprise

Team adds central billing, SSO, admin controls, enterprise search, and no model training on your content by default. Enterprise adds user and organization spend limits, role-based access, SCIM, audit logs, a Compliance API, custom data retention and a HIPAA-ready option.

How do Claude usage limits work?

Claude usage limits are counted in sessions that reset every five hours. Anthropic asks you to think of it as a "conversation budget": how many messages you can send, or how long you can work in Claude Code, before the session allowance runs out. Anthropic's usage limits article names four things that decide how fast it goes:

• the length and complexity of your conversations,
• the features you use,
• the model you are chatting with,
• the effort level you have selected.

On Max there is a second limit on top: a weekly usage limit across all models. It resets at a fixed day and time assigned to your account, not a week after you subscribed, and you get the full allowance each cycle. The next reset time is in Settings > Usage.

When you hit a limit you have three options: wait for the reset, upgrade the plan, or pay for usage credits, which paid plans can buy to keep working past the included allowance.

Usage limits are different from length limits. A usage limit is how much you can do over time; a length limit is how long a single conversation can get. For a length limit, start a new conversation or move the material into a project.

Where does the allowance actually go?

My own Claude Code statistics over several months of daily work answer this better than any plan page. About 99% of all tokens were cache reads: the model re-reading context it had already loaded. The text Claude wrote back was about a quarter of one percent, and what I typed was a rounding error. That is why a long session drains a limit much faster than the length of the answers suggests: every step re-reads everything before it. Anthropic lists the model as one of the factors, and on the API an Opus cache read costs 2.5 times a Sonnet 5 one ($0.50 against $0.20 per million tokens), which matches the jump I saw when Opus became my default.

What that means in practice:

• Start a new session when the task changes. A fresh context is cheaper than dragging an old one along.
• Use Sonnet for routine work and keep Opus for the problems that need it.
• Watch the usage screen, not your gut: the limit is spent by context, not by how much you asked.

I have had to slow down to stay inside the limit more than once. In the session logs I still have, from late August to mid-September, no session was cut off by a limit.

Is Claude Max worth it over Pro?

Claude Max is worth it when Pro's five-hour limit interrupts your work several times a week, not before. The arithmetic per unit of usage:

Max 5x is not a discount. It costs exactly five Pro subscriptions for five times the usage. What you buy is headroom in one account, priority access when Claude is busy, higher output limits, and Fable within half of your weekly limit. Max 20x is the tier that actually gets cheaper per unit, at half of Pro's price per unit of usage. The rule of thumb: if Pro runs out regularly, go to Max 5x; if Max 5x runs out regularly, go to Max 20x.

The multiples are Anthropic's description of per-session usage, not a measured number of messages, and Max also has the weekly cap that Pro's comparison does not show.

ChatGPT plans and prices (official, September 2026)

OpenAI publishes how many Codex messages fit into one five-hour window on each plan. Each cell is a range, not a fixed number: "5-45" means somewhere between 5 and 45 messages before the window runs out. OpenAI says the count depends on the model, the size and complexity of the task, and whether it runs locally or in the cloud: small scripts or routine functions use a fraction of the allowance, larger projects and long-running tasks use much more.

GPT-6 Astra gets the smallest allowance on every plan: on Plus, as few as 5 messages in a five-hour window. OpenAI says these are estimates rather than fixed limits, that local messages and cloud chats share one allowance, and that weekly limits may also apply. Source: OpenAI's pricing documentation, checked 12 September 2026. The consumer page chatgpt.com/pricing refused automated access that day, so check it for anything plan-specific beyond Codex.

Claude or ChatGPT: which plan costs less?

At list price, neither. Both start paid plans at $20 a month and both sell 5x and 20x tiers at $100 and $200. The real differences are elsewhere:

• Cheapest paid entry: ChatGPT Go at $8. Claude has nothing between Free and Pro.
• Annual billing: Claude Pro is $17 a month billed annually. Claude Max is monthly only.
• How usage is described: Anthropic gives multiples of Pro per session plus a weekly cap on Max. OpenAI gives message ranges per five-hour window per model, plus weekly limits that "may apply".

Because the two companies count usage in different units, the same $20 does not buy the same amount of work, and neither publishes a number you can compare directly. That comparison needs a measurement: the same tasks, run on both, until each plan stops. I have not run it yet, so this page does not claim a winner.

Change log

Frequently asked questions

What are the Claude plans in 2026?

Claude has four personal plans: Free ($0), Pro ($20 a month, or $17 a month billed annually at $200), Max 5x ($100 a month) and Max 20x ($200 a month). For organizations there is Team, with standard seats at $25 a month ($20 annually) and premium seats at $125 ($100 annually), and Enterprise at $20 a seat plus usage at API rates.

What is the 5-hour limit in Claude?

Paid Claude plans measure usage in sessions, and the session allowance resets every five hours. How far a session goes depends on conversation length, the features you use, the model and the effort level. When you hit it, you wait for the reset, upgrade, or spend usage credits.

Does Claude Max have a weekly limit?

Yes. On top of the five-hour sessions, Max plans have a weekly usage limit that applies across all models. It resets at a fixed day and time assigned to your account, not from when you subscribed, and you can see the next reset in Settings > Usage.

Is the $100 Claude Max plan worth it?

Max 5x costs five times Pro and gives five times Pro's usage per session, so it is not cheaper per unit: you pay for headroom and priority access at busy times. Max 20x is the plan that gets cheaper per unit, at twice Pro's rate of usage per dollar. Max 5x is worth it if Pro's limit interrupts your work several times a week.

Is Claude cheaper than ChatGPT?

At list price the ladders match: both start paid plans at $20 a month and both offer 5x and 20x tiers at $100 and $200. ChatGPT also has an $8 Go plan below that, and Claude Pro drops to $17 a month on annual billing. The two companies count usage differently, so equal prices do not mean equal work.

How much does ChatGPT cost in 2026?

OpenAI lists ChatGPT Free at $0, Go at $8 a month, Plus at $20, Pro at $100 (5x Plus's rate limits) or $200 (20x), and Business at $20 per user per month. For Codex, OpenAI estimates usage on Plus, Pro and Business in messages per five-hour period, and says weekly limits may also apply.