Claude Max for companies: cost, limits, DPA
Claude Max costs 100 or 200 USD per month and stays a consumer plan without a DPA. When Team, Enterprise or Bedrock EU is the better company choice.
TL;DR
Claude Max is a consumer subscription for one person: 100 USD (5x) or 200 USD (20x) per month, billed monthly, with no DPA, no SSO and no central billing. For company data we recommend Team as the floor. If inference must stay in the EU, run Claude on Bedrock with eu.anthropic.claude-sonnet-4-6.
What does Claude Max cost in 2026?
Claude Max costs 100 USD for the 5x tier and 200 USD for 20x, per month, excluding tax. There is no annual Max plan. Pro sits below it at 17 USD per month on annual billing (200 USD up front) or 20 USD billed monthly. The public Max card only says “From $100”. The 200 USD figure comes from the help article.
The Max help article states both tiers and limits them to web subscriptions. Mobile prices through the App Store or Google Play may differ. claude.com/pricing, read on 30 September 2026, repeats the “From $100” card and the monthly-only billing cycle.
| Plan | Web price (excl. tax) | Billing | Usage vs Pro |
|---|---|---|---|
| Pro | 20 USD monthly, 17 USD annual | month or year | 1x (baseline) |
| Max 5x | 100 USD | month only | 5x Pro per 5-hour window |
| Max 20x | 200 USD | month only | 20x Pro per 5-hour window |
| Team Standard | 25 USD monthly, 20 USD annual | month or year | 1.25x Pro, per person |
| Team Premium | 125 USD monthly, 100 USD annual | month or year | 6.25x Pro, per person |
| Enterprise | 20 USD per seat plus API usage | year only | no included seat pool |
Team prices and multipliers are in the Team help article. Team covers 2 to 150 people. Enterprise is 20 USD per seat plus usage at API rates. Every listed price excludes tax.
One developer on Max 5x pays 1,200 USD a year before tax. The same person on Team Premium at annual billing also costs 1,200 USD, but under a company contract with SSO. Five developers on private Max 5x accounts cost 6,000 USD and remain five separate consumers. The same five on Team Premium also cost 6,000 USD, plus at least one extra Standard seat because Team needs two members.
The full plan grid from Free to Enterprise is in our Claude pricing plans 2026 overview. Claude Code is in every paid plan and shares the usage pool with chat, per the Pro and Max Claude Code article.
What limits does Claude Max actually have?
Every plan uses a rolling five-hour session window. Paid plans add a weekly cap. Web, desktop, mobile and Claude Code draw from one pool. There is no fixed message count. Max 5x is five times Pro per window. Max 20x is twenty times Pro. Anthropic may also apply weekly or monthly caps, or model and feature limits.
The Max help article says the session limit resets every five hours. The weekly limit applies across models and resets on a fixed weekday assigned to the account. You can see the next reset under Settings, Usage. The pricing FAQ adds: when you hit a limit you wait, change plan, or, on paid plans, turn on usage credits at standard API rates.
| Limit | Free / Pro | Max 5x | Max 20x | Team Standard | Team Premium |
|---|---|---|---|---|---|
| Window | 5 hours, rolling | 5 hours | 5 hours | 5 hours, per person | 5 hours, per person |
| Amount vs Pro | Pro at least 5x Free | 5x Pro | 20x Pro | 1.25x Pro | 6.25x Pro |
| Weekly cap | yes on paid plans | yes | yes | yes, per person | yes, per person |
| Fable | Pro: credits | 50% of weekly limits | 50% of weekly limits | 50% on Premium seats | 50% of weekly limits |
| Priority at peak | no on Pro | yes | yes | yes | yes |
| Org spend controls | no | no | no | yes | yes |
Max includes Fable, Opus, Sonnet and Haiku. The pricing table lists context as “Up to 1M varies by model”. Max has no user-level or organisation spend controls and no project sharing. Incognito chats exist and are not used to improve Claude, even if model improvement is on.
When the pool is empty, Claude Code may offer Console API credits. The help article warns that an ANTHROPIC_API_KEY environment variable bypasses the subscription and bills tokens. To stay on the plan, decline the API option or log in with the Max account only.
Why Claude Max has no company DPA
Max is a consumer product. The DPA help article covers commercial products such as Claude for Work and the API. It points consumer products the other way. The Max help article opens by saying the plan is for individual consumers. There is no processor agreement for Max.
The Anthropic DPA (effective 24 February 2025) is built into the Commercial Terms. Anthropic is the processor, the customer is the controller. For transfers outside the EEA it uses the Standard Contractual Clauses, Modules Two and Three, under Implementing Decision (EU) 2021/914. You accept that DPA with Team, Enterprise or the API. A private Max account does not.
Retention on consumer plans, from the Privacy Center article dated 1 July 2026:
- Deleted chats leave the history immediately and leave back-end storage within 30 days.
- If you allow training, Anthropic may keep de-identified data in training pipelines for up to 5 years.
- Usage Policy flags: inputs and outputs up to 2 years, classification scores up to 7 years.
- Feedback (thumbs, bug reports) is kept for 5 years.
On Max, model training is opt-out. Team and Enterprise turn training off by default. For company data that is the real split, not model quality.
Commercial products store data in the US. Traffic may go to selected countries in the US, Europe, Asia and Australia, says the server location article. First-party data residency accepts only global and us for inference_geo. There is no EU value. The contract path across Bedrock, Vertex and Foundry is in our Claude GDPR comparison.
Claude Max vs Team: what companies actually buy
Max buys capacity. Team buys administration. Max has no SSO, no domain capture, no central billing, no admin controls for connectors, no roles, no SCIM, no audit logs, no Compliance API and no custom retention. The pricing comparison sets every one of those rows to No for Max 5x and Max 20x.
Team needs at least two members and stops at 150 seats. Standard seats sit near Pro on price. Premium seats sit near Max 5x, under a company contract. Limits are per member, not for the team as a whole. One person at the cap does not block the others. That is in the Team help article.
| Criterion | Max 5x / 20x | Team | Enterprise |
|---|---|---|---|
| Contract | Consumer | Commercial Terms plus DPA | Commercial Terms plus DPA |
| Account owner | the person | the organisation | the organisation |
| SSO / domain | no | yes | yes, plus domain capture |
| SCIM, audit logs, custom retention | no | no | yes |
| Training | opt-out | none by default | none by default |
| Payment | personal card | central card | card, ACH, invoice via sales |
| EU inference | no | no | US-only at 1.1x, no EU |
| Minimum size | 1 | 2 | 20 self-serve, 50 via sales |
Our position: a private Max account for customer data, source code with personal logs, or HR text is a governance mistake. The account belongs to the employee. When they leave, the history leaves with them, or it stays in a private Anthropic account the company cannot control. Team fixes that for chat and Claude Code. Enterprise adds SCIM and audit. No Anthropic plan gives you EU inference. That decision lives in Claude Enterprise vs Bedrock.
How we would choose:
- Classify prompts. Do the five most common prompts contain personal data, or only public code?
- Name the account owner. Does the subscription belong to the company or the person?
- Check identity. Do you need SSO on first login, and later SCIM to revoke access?
- Size the pool. Is 1.25x Pro (Team Standard), 5x (Max 5x or Team Premium) or 20x (Max 20x) enough?
- Decide residency. Must inference stay in the EU? Then Max and Team both drop out.
Alternative: Bedrock EU instead of private Max accounts
Claude Max runs on Anthropic infrastructure with no EU switch. Amazon Bedrock keeps inference inside the EU with a geographic profile. eu.anthropic.claude-sonnet-4-6 can start in Frankfurt and route to Stockholm, Milan, Spain, Ireland and Paris. That is on the Claude Sonnet 4.6 model card.
Geographic cross-Region inference says requests to a geography-tied profile stay in that geography. Data stays in the source Region by default. Prompts and outputs may move inside the geography. Abuse-detection storage, where it exists, sits in the destination Region.
Platform pricing applies a 10 percent premium on regional and multi-region endpoints versus global endpoints from Claude Sonnet 4.5, Haiku 4.5 and Opus 4.5 onward. Global endpoints have no residency guarantee. On the first-party API, US-only inference is 1.1x for models from 4.6 onward.
List prices read on 30 September 2026 at platform.claude.com:
| Model | Input / output per 1M tokens |
|---|---|
| Claude Sonnet 5.5 | 2 / 10 USD |
| Claude Opus 5.5 | 4 / 20 USD |
| Claude Fable 5.1 | 10 / 50 USD |
| Claude Sonnet 5 | 2 / 10 USD (now the standard price) |
A developer who burns 20 million input and 1 million output tokens a month on Sonnet 5.5 pays 50 USD on the API. That undercuts Max 5x if the load really looks like that, and loses if sessions are shorter. Bedrock is not cheaper by default. It is cheaper when you need an AWS DPA, IAM, CloudTrail and an EU route.
Call from Frankfurt. Do not copy the global profile from the AWS sample:
import boto3
client = boto3.client("bedrock-runtime", region_name="eu-central-1")
response = client.converse(
modelId="eu.anthropic.claude-sonnet-4-6",
messages=[
{
"role": "user",
"content": [{"text": "Summarise the internal policy in five sentences."}],
}
],
)
print(response)IAM and Claude Code setup is in our Claude Code on Bedrock EU guide. Regions and prices sit on the Amazon Bedrock EU hub.
FAQ
How much does Claude Max cost per month?
Max 5x costs 100 USD. Max 20x costs 200 USD. Both are monthly and exclude tax. There is no annual Max plan. Those figures are for the web subscription. Mobile prices may differ.
Claude Max vs Team: what is the difference?
Max buys more capacity for one person. Team buys administration for 2 to 150 people. Team adds SSO, central billing, no training by default and the commercial DPA. Max adds 5x or 20x Pro usage and priority at peak load, and none of the admin row. For company data, Team is the floor.
Does Claude Max include a DPA?
No. Anthropic’s DPA covers commercial products. The help article names Claude for Work and the API. Free, Pro and Max are consumer products. For a processor agreement you need Team, Enterprise, the API, or a cloud marketplace such as Bedrock.
Is Claude Code included in Claude Max?
Yes. Claude Code is in every paid plan and shares the pool with chat. IDE use in VS Code, Cursor and JetBrains counts against the same pool. An ANTHROPIC_API_KEY bills tokens instead of the subscription.
Does Claude Max keep data in the EU?
No. Anthropic does not document EU residency for consumer products. Commercial products store data in the US. First-party inference has no EU option, only global and us. For EU processing use Bedrock with an eu. profile or Google Cloud.
Is Max 20x cheaper than five Team Premium seats?
Max 20x at 200 USD sits near two annual Team Premium seats (200 USD), not five. Five Premium seats at annual billing cost 500 USD a month. Max 20x is one private super-pool. Team Premium spreads 6.25x Pro across named seats under a company contract. Five developers belong on Team, not on five Max accounts.
Sources
- Claude: Pricing (30 September 2026)
- Claude Help Center: What is the Max plan? (30 September 2026)
- Claude Help Center: What is the Team plan? (30 September 2026)
- Claude Help Center: Using Claude Code with Pro or Max (30 September 2026)
- Claude Help Center: How do I view and sign your DPA? (30 September 2026)
- Anthropic Data Processing Addendum (30 September 2026)
- Anthropic Privacy Center: How long do you store my data? (30 September 2026)
- Anthropic Privacy Center: Where are your servers located? (30 September 2026)
- Anthropic Docs: Pricing (30 September 2026)
- Anthropic Docs: Data residency (30 September 2026)
- AWS Bedrock: Claude Sonnet 4.6 model card (30 September 2026)
- AWS Bedrock: Geographic cross-Region inference (30 September 2026)