Claude Pricing: Free vs Pro, Max, Team, Enterprise and API Costs
GuidesThursday, 06 August 2026 at 13:00
Claude is free to try. Paid individual access starts with Claude Pro at $20 per month, while Claude Max starts at $100 per month. Team uses per-seat pricing, Enterprise combines a seat fee with metered usage, and the Claude API is charged separately according to the model and number of tokens processed.
The headline prices are easy. The difficult part is understanding what is being purchased. A Claude subscription provides access to a collection of products within usage limits. The API charges for model consumption. Claude Code can use either route. Team and Enterprise add administration and commercial data terms, not merely more messages.
This guide compares Anthropic’s current published pricing and explains which plan fits which workload. For the broader product map, start with our complete Claude guide.
Claude prices at a glance
Prices below are Anthropic’s standard US list prices before applicable tax on July 27, 2026. Local currency, app-store billing, promotions and contract terms can differ.
| Plan | Published price | Intended user | Main reason to choose it |
| Free | $0 | New or occasional user | Test Claude on real tasks |
| Pro | $20 monthly or $200 annually | Regular individual user | More usage and access to paid productivity features |
| Max 5x | $100 monthly | Individual power user | Five times the Pro usage allowance |
| Max 20x | $200 monthly | Very heavy individual user | Twenty times the Pro usage allowance |
| Team standard seat | $25 monthly or $20 per month with annual billing | Managed teams of 2–150 | Central billing, SSO, enterprise search and no training by default |
| Team premium seat | $125 monthly or $100 per month with annual billing | Heavy users inside a Team workspace | Five times the standard Team-seat usage |
| Enterprise | $20 per seat per month plus usage at API rates, annual billing | Large or governed organizations | Fine-grained permissions, SCIM, audit and retention controls |
| API | Usage-based | Developers and products | Direct model access without a consumer subscription |
Annual individual Pro is billed as $200 upfront, equivalent to about $16.67 per month. Team annual prices are expressed per seat per month but normally involve an annual commitment.
“Five times” and “twenty times” do not mean a fixed number of messages. Anthropic applies usage limits, and consumption depends on the selected model, conversation length, attachments, tools and task. A long Claude Code session can use far more capacity than a short chat.
Is Claude free?
Yes. Claude Free costs nothing and provides access to the core assistant on the web, desktop and mobile within limits. Anthropic’s current plan page lists capabilities such as web search, memory, file creation, code execution, connectors, extended or adaptive thinking and ordinary writing and analysis, but plan availability and allowances can change.
Free is the correct starting plan for most people. Test tasks that resemble the work you would actually pay to perform:
- edit a real draft;
- compare two non-sensitive documents;
- research a current question and inspect the citations;
- create a small spreadsheet or presentation;
- ask Claude to explain a piece of code;
- or build an Artifact.
The free plan is not a predictable production resource. Availability can be tighter at busy times, model choice is more limited and advanced work can exhaust the allowance. If a task must finish at a particular time, a paid subscription or metered API is easier to plan around.
When Free is enough
Stay on Free when Claude is an occasional second opinion, the interruptions do not matter and no paid-only feature is essential. A free user who asks a few short questions each week gains little from buying theoretical extra capacity.
When to leave Free
Upgrade when one of these statements is consistently true:
- useful work is regularly stopped by the allowance;
- Projects, Research, Cowork, Claude Code or another paid capability is part of the workflow;
- the time saved is worth more than the subscription;
- or the data belongs in a managed commercial workspace rather than a personal account.
The last reason is different from the first three. Buying Pro does not turn a consumer account into an organizational data-governance product.
Claude Pro: $20 monthly or $200 annually
Claude Pro is Anthropic’s main paid individual plan. It includes everything in Free and increases usage. Anthropic currently lists Claude Code, Claude Cowork, Claude Design, Claude Science, unlimited Projects, Research, additional model access and Claude for Microsoft 365 among the Pro benefits.
The annual price saves $40 against twelve monthly payments:
- monthly for one year: $20 × 12 = $240;
- annual payment: $200;
- annual saving: $40.
Annual billing is economical only if you expect to use Claude for most of the year. Monthly billing is more flexible when comparing Claude with ChatGPT, Gemini or another fast-changing product.
Who should buy Pro?
Pro is the best default for an individual who uses Claude several days a week for writing, research, documents or moderate coding. At $20, it needs to save only a small amount of professional time each month to pay for itself.
It is less convincing for a user who wants unlimited heavy Claude Code or long-running agent work. “Includes Claude Code” means access within plan limits, not an unmetered engineering service.
What can still interrupt a Pro user?
Usage is influenced by:
- the model selected;
- the amount of text already in the conversation;
- uploaded files;
- reasoning effort;
- Claude Code and Cowork activity;
- tool use;
- and overall demand.
Starting a new chat after the old context is no longer needed can conserve capacity. Keeping a 100-page source and dozens of irrelevant turns in every request makes the model repeatedly process more material.
Claude Max: $100 or $200 per month
Claude Max is for individual power users. It has two levels:
- Max 5x at $100 per month;
- Max 20x at $200 per month.
Both include the Pro feature set. The purchase is mainly for a larger usage allowance, higher output limits, priority access during busy periods and earlier access to selected features.
Max 5x versus Max 20x
Max 5x is the logical step when Pro is demonstrably too small and the user can identify the valuable work being interrupted. Max 20x is aimed at people who depend on Claude throughout the day for coding, research, document production or agentic tasks.
The price jump should be tested against alternatives:
- Could an API workflow be cheaper for a repeatable task?
- Could Sonnet or Haiku handle part of the workload?
- Is old conversation context consuming capacity?
- Does a Team or Enterprise account better match company use?
- Would two specialized products produce better results than one maximum subscription?
Do not infer that Max 20x is “unlimited.” Anthropic still applies usage limits, and model-specific or weekly restrictions can exist.
Is Max worth it?
Max is worth it when the marginal work completed between Pro and Max has a value above $80 or $180 per month respectively. For a developer billing professional hours, that can be easy. For casual personal use, it usually is not.
Track interrupted sessions for a week before upgrading. Record the task, estimated value and whether a cheaper model or fresh chat would have solved the problem. Buy capacity from evidence, not frustration after one unusually large prompt.
Claude Team pricing
Claude Team is for organizations with 2 to 150 users. It offers two seat types that can be mixed:
| Team seat | Annual billing | Monthly billing | Relative usage |
| Standard | $20 per seat/month | $25 per seat/month | More than individual Pro |
| Premium | $100 per seat/month | $125 per seat/month | Five times a standard seat |
Team currently includes Claude Code, Cowork, Design, Science, Microsoft 365 connections, enterprise search, central administration, SSO, connector controls and enterprise desktop deployment. Anthropic says Team content is not used for model training by default.
Why Team is not just a group discount
An organization buys Team for a managed boundary:
- company-owned billing;
- member and access management;
- SSO;
- administrative controls;
- shared organizational features;
- and commercial data treatment.
If employees are using personal Pro accounts for company documents, adding everyone to Team is not merely a cost comparison. It changes ownership, administration and policy enforcement.
Standard and premium seats
Mixing seats can reduce cost. An employee who uses Claude for two meeting briefs per week does not need the same allowance as an engineer or analyst running agentic work all day.
Start with roles and measured use. Assign premium seats to people whose capacity limits block valuable work, then review. A prestige-based allocation creates cost without productivity.
Claude Enterprise pricing
Anthropic’s current self-serve Enterprise offer is priced at $20 per seat per month plus usage charged at API rates. It is billed annually. A sales-assisted Enterprise agreement can have tailored commercial terms, commitments and support.
This structure is different from a traditional flat-price “unlimited” enterprise seat. The seat provides access and management; actual model use creates a variable cost.
Enterprise adds capabilities such as:
- role-based access;
- SCIM provisioning;
- audit logs;
- Compliance API access;
- organizational spend controls;
- custom data-retention settings;
- network-level access controls;
- IP allowlisting;
- and a HIPAA-ready offering where configured and eligible.
The listed seat price is therefore not the expected total cost. A realistic budget needs:
total cost = seat fees + model usage + implementation + governance + support and integration work
Agentic tasks can use many model and tool calls. Pilot the exact workflow and collect usage data before projecting an enterprise-wide bill.
Team or Enterprise?
Choose Team when SSO, central administration and managed collaboration cover the requirements. Choose Enterprise when the organization needs fine-grained roles, automated provisioning, audit, custom retention, spend controls, network restrictions or negotiated compliance and support.
The choice should come from the security and procurement requirements, not company size alone.
Does Claude Pro include Claude Code?
Yes. Anthropic currently includes Claude Code in Pro, Max and eligible workplace plans. You can also use Claude Code through a Claude Console account and pay according to API consumption.
These routes behave differently:
- Subscription access draws from the plan’s usage allowance.
- Console access produces a metered API bill.
- Team and Enterprise add organizational administration.
- Supported cloud providers have their own billing and identity arrangements.
An occasional developer may prefer the predictable Pro price. A production team may prefer metered billing, spend controls and a cloud architecture. Read our complete Claude Code guide before choosing solely on price.
Claude API pricing
The Claude API charges separately from Claude subscriptions. One million tokens is written as MTok. Input tokens include the prompt and context sent to the model; output tokens are generated by the model.
Current model rates
| API model | Input per MTok | Output per MTok | Prompt cache write | Prompt cache read |
| Claude Fable 5 | $10 | $50 | $12.50 | $1 |
| Claude Opus 5 | $5 | $25 | $6.25 | $0.50 |
| Claude Sonnet 5 | $2 introductory | $10 introductory | $2.50 introductory | $0.20 introductory |
| Claude Haiku 4.5 | $1 | $5 | $1.25 | $0.10 |
Anthropic says Sonnet 5’s introductory price of $2 per million input tokens and $10 per million output tokens runs through August 31, 2026. The published standard price after that date is $3 and $15. Any calculator or article should be updated when that transition occurs.
The cache-write rates shown use the standard five-minute time-to-live. Longer caching can use different pricing. US-only inference is available for eligible workloads at 1.1 times the normal input and output rate.
Simple API cost example
Suppose an application sends 100,000 input tokens and receives 20,000 output tokens with Sonnet 5 at its introductory rates:
- input: 0.1 MTok × $2 = $0.20;
- output: 0.02 MTok × $10 = $0.20;
- model cost: $0.40.
That does not automatically include web search, managed-agent runtime, extra code-execution hours or other paid tools.
Why output length matters
Output tokens are more expensive than input tokens for every current model. A vague instruction to produce an extremely long response can cost more than sending a substantial source document.
Control output with a useful specification:
- the required fields;
- a maximum length;
- a schema;
- a stopping condition;
- or a concise first pass followed by targeted expansion.
Do not force brevity when completeness is the product requirement. The goal is to remove unused output, not useful content.
Additional platform costs
Anthropic’s current platform pricing includes:
- managed-agent runtime at $0.08 per active session-hour, in addition to normal token rates;
- web search at $10 per 1,000 searches, excluding the tokens used to process results;
- code execution with 50 free hours per organization per day and additional time at $0.05 per container-hour;
- Fast mode for Opus 5 at $10 per MTok input and $50 per MTok output;
- and regional or service-tier modifiers where applicable.
Message Batches can reduce model token charges by 50 percent for asynchronous workloads. Prompt caching can reduce the repeated cost of a stable prefix. The two discounts can stack, although cache hits in concurrent batch traffic are not guaranteed.
Our Claude API tutorial explains requests, token use, tools, retries and cost controls. The Claude models guide helps select the lowest-cost model that still meets the quality requirement.
How to estimate a monthly API bill
Use measured workload data:
- Choose a representative set of requests.
- Record input and output tokens.
- Separate cached and uncached input.
- Count tool calls and runtime charges.
- Multiply by expected monthly volume.
- Add retries, failed runs and growth headroom.
- Compare at least two models.
A useful formula is:
monthly model cost = (input MTok × input rate) + (output MTok × output rate) + cache and tool charges
Do not multiply one perfect demonstration by projected traffic. Production prompts are messier, users request longer outputs and agent loops can retry. Use percentiles, not only the average.
Which Claude plan should you choose?
Choose Free if:
- you are testing Claude;
- use is occasional;
- interruptions are acceptable;
- and the information is appropriate for a personal consumer account.
Choose Pro if:
- you use Claude regularly as an individual;
- $20 is justified by time saved;
- you want paid productivity features;
- and moderate Claude Code or Cowork use is enough.
Choose Max 5x if:
- Pro limits repeatedly stop valuable work;
- the workload is intensive but still individual;
- and $100 is cheaper than the lost time.
Choose Max 20x if:
- Claude is a central daily work environment;
- you run long coding, research or agentic sessions;
- and you have measured a real need for the highest individual allowance.
Choose Team if:
- the organization needs central billing and administration;
- employees need a commercial workspace;
- SSO and connector controls are sufficient;
- and the company has no requirement for the advanced Enterprise controls.
Choose Enterprise if:
- SCIM, audit logs, custom retention, role-based access or network controls are requirements;
- procurement needs negotiated terms;
- or variable, governed usage is more appropriate than consumer allowances.
Choose API billing if:
- Claude is inside a product or automated workflow;
- usage needs to be measured per request;
- the application needs tools or structured output;
- or a developer wants direct model control.
One person can legitimately use both a subscription and the API. The subscription supports interactive work; the API powers repeatable software.
Claude versus ChatGPT pricing
Claude Pro and ChatGPT Plus both use a $20 monthly entry point in the United States. That does not make the plans equivalent. Each bundles different models, products and allowances.
Compare what you will actually use:
- Claude Projects, Artifacts, Research, Cowork and Claude Code;
- versus ChatGPT Projects, Deep Research, image creation, custom GPTs, plugins, ChatGPT Work and Codex;
- plus the relevant memory, file, voice, connector and business features.
The correct comparison is not “which company gives more AI?” It is “which subscription completes more of my tasks with less correction and acceptable data handling?” Our Claude versus ChatGPT guide provides that product-level comparison. AI World Today also maintains a separate ChatGPT pricing guide.
Hidden costs to include
The subscription or token bill is only one cost.
Review time
If a draft takes ten minutes to generate and forty minutes to repair, the model did not save forty minutes. Measure the complete workflow.
Failed agent runs
An agent may choose the wrong tool, loop, retry or generate an unusable file. Those actions consume tokens and staff attention.
Integration
API authentication, monitoring, evaluation, data pipelines and user interfaces require engineering. A low model price does not make the application free.
Governance
Security review, vendor management, employee training, incident response and audit can be appropriate costs, especially when Claude can access internal systems.
Switching and duplicate subscriptions
Paying for several overlapping AI assistants can be sensible during evaluation and wasteful afterward. Review active subscriptions quarterly and assign an owner.
Frequently asked questions
How much is Claude Pro?
Claude Pro costs $20 per month or $200 billed annually in the United States before tax.
How much is Claude Max?
Max 5x costs $100 per month and Max 20x costs $200 per month. Both remain subject to usage limits.
Can I use Claude Code for free?
Claude Code requires an eligible subscription or consumption-based account for normal use. Temporary promotions or limited access can change. The core Claude Free plan should not be treated as an unlimited Claude Code plan.
Is the Claude API included with Pro?
No. API usage is billed separately. Claude Code can authenticate through an eligible Claude subscription or through developer and cloud routes, but ordinary API calls do not become free because you pay for Pro.
Does Team cost less than Pro?
A standard Team seat can cost the same as Pro with annual billing, but Team normally involves an organizational workspace, seat requirements and annual terms. It is not simply a cheaper personal subscription.
Is Claude Enterprise unlimited?
No. Anthropic’s current self-serve Enterprise model combines a per-seat fee with usage at API rates. Sales-assisted contracts can have different commitments, but organizations should model variable usage.
Which API model is cheapest?
Claude Haiku 4.5 has the lowest current base token price. It is not automatically cheapest for every workflow: a stronger model may complete a difficult task in fewer retries. Test total cost per successful result.
Can Claude pricing change?
Yes. Models, plan benefits, limits and promotional rates change. Sonnet 5 already has a published introductory-price end date of August 31, 2026. Verify Anthropic’s official price page before a purchasing decision.
Bottom line
Claude Free is for testing, Pro is the sensible default for regular individual work, and Max is capacity for measured power users. Team and Enterprise solve organizational administration and governance needs. The API is a separate consumption product.
The cheapest plan is not the one with the lowest price. It is the one that produces the required result at the lowest complete cost, including review, failed runs, integration and risk. Start small, measure real work and upgrade only when the evidence shows where the constraint is.
byDavid Porter
Loading