Claude pricing is not one price. A personal Claude subscription, a Team seat, Claude Code authenticated through a plan, Claude Code using an API key, and a production API workload can all generate different bills. The cheapest-looking number is useless until you know which route your work actually uses.
The figures below come from Anthropic’s public pricing, help, and API-limit pages checked August 18, 2026. The most important rule is simple: Claude app subscriptions and Claude API token charges are separate budgets. Enterprise also combines seats with metered usage, so its public seat price is not the final total.
If your work also needs GPT, Gemini, Perplexity, image, or video tools, GlobalGPT offers a different route: one multi-model workspace. That can simplify everyday research, writing, planning, and content production, but it should not be described as an official Claude API, Claude Code, or Enterprise replacement.
Five different ways Claude-related access is billed
Start by identifying the billing system before comparing prices. A $20 subscription and a $2-per-million-token input rate do not compete for the same job, while Enterprise adds usage on top of its seat entry price.
| Route | What you buy | Billing unit | Do not confuse it with |
|---|---|---|---|
| Claude individual app | Claude on web, desktop, and mobile; plan features can include Claude Code | Free, monthly, or annual subscription | General API token credits |
| Team / Enterprise | Seats, admin and organization features; Enterprise can add metered usage | Per seat, with Enterprise usage at API rates | A single fixed enterprise total |
| Claude Code | Coding workflow using a plan allowance or API credentials | Shared plan capacity or API token billing | A separate unlimited coding subscription |
| Claude API | Programmatic model calls for apps, agents, and internal tools | Input, output, cache, and tool units | Claude Pro or Max monthly fees |
| GlobalGPT | A separate multi-model workspace with platform credits | GlobalGPT monthly or annual subscription | Anthropic tokens, Claude plan usage, or official enterprise controls |
For a deeper feature-by-feature comparison of personal and organization subscriptions, use the separate Claude AI plans guide. This page keeps its main focus on the pricing boundaries and API economics.
Claude Free, Pro, and Max pricing
Free is the evaluation tier. Pro is the practical starting point for regular individual work. Max is for people whose paid work is repeatedly interrupted by Pro capacity; it is not a modest feature add-on.
| Plan | Public price | Best fit | Buying boundary |
|---|---|---|---|
| Free | $0 | Occasional use and evaluation | Usage is limited. |
| Pro | $20/month or $200/year | Regular individual work | The annual price is prepaid; the page displays a $17 monthly equivalent. |
| Max 5x | $100/month | People who repeatedly hit Pro capacity | 5x Pro capacity per session; no verified annual amount is published in this guide. |
| Max 20x | $200/month | Heavy daily Claude and Claude Code use | 20x Pro capacity per session; usage still has plan limits. |
Monthly price ladder · USD · checked August 18, 2026
Max is a capacity purchase, not a small step above Pro
Source: Claude pricing and Choose a Claude plan.
Pro’s annual option is clear: $200 prepaid, displayed as $17 per month. Max is different. Until Anthropic publishes a consistent annual amount across its pricing and help pages, use the verified $100 and $200 monthly figures and confirm the checkout terms before buying.
Claude Team and Enterprise pricing
Team pricing is per seat and splits capacity into Standard and Premium seats. Enterprise requires another layer of budgeting because the self-serve entry combines a seat charge with usage at API rates.
| Plan | Public price | Capacity / billing note | Budget implication |
|---|---|---|---|
| Team Standard | $20/seat/month annually; $25 monthly | For teams of 2–150 | Annual billing saves $5 per seat each month before taxes. |
| Team Premium | $100/seat/month annually; $125 monthly | 5x Standard-seat usage | Choose it for capacity, not because it is a feature-equivalent personal Max plan. |
| Enterprise self-serve | $20/seat + usage at API rates | Annual billing shown | Seat price is only one part of the total. |
| Enterprise sales-assisted | Contact sales | Contract terms vary | Confirm seats, usage, support, compliance, and incentives directly. |
For Team, annual billing reduces the displayed monthly seat rate, but the real decision is whether every member needs Premium capacity. For Enterprise, ask sales or procurement to confirm seat minimums, contract terms, usage, support, compliance, and incentives. Do not multiply $20 by seat count and call that the enterprise total.
Claude Code: subscription allowance vs API billing
Claude Code is included in Pro and Max, but that sentence needs a boundary. Activity in Claude Code, supported IDEs, and the Claude app draws from the same plan allowance. A long coding session can therefore affect the capacity left for web or desktop work.
Claude Code has two billing paths
Plan authentication
Pro or Max can cover Claude Code, but the terminal, IDE, web, desktop, and mobile experiences draw from the same plan allowance.
API authentication
If ANTHROPIC_API_KEY is set, Claude Code uses API billing instead of the subscription allowance. API credits are charged at standard API rates.
This is also why a Claude Code cost estimate should begin with authentication. If the terminal is using your plan, compare Pro and Max capacity. If it is using an API key or API credits, estimate tokens and tool charges. The practical setup guide on using Claude AI for coding is a useful companion, while the Claude vs ChatGPT coding comparison helps when model behavior matters as much as price.
Claude API pricing by latest model
Anthropic lists Fable 5, Opus 5, Sonnet 5, and Haiku 4.5 as its latest models on the pricing page. Opus 4.8 and Sonnet 4.6 appear under legacy models, so they should be treated as migration or compatibility choices rather than the headline current lineup.
| Latest model | Input / MTok | Output / MTok | 5-min cache write | Cache read |
|---|---|---|---|---|
| Claude Fable 5 | $10 | $50 | $12.50 | $1.00 |
| Claude Opus 5 | $5 | $25 | $6.25 | $0.50 |
| Claude Sonnet 5 | $2 | $10 | $2.50 | $0.20 |
| Claude Haiku 4.5 | $1 | $5 | $1.25 | $0.10 |
Output price · USD per MTok · checked August 18, 2026
Long answers amplify the price gap
Source: Claude API pricing. Token charges are separate from Claude app subscriptions.
The model label alone is not a cost estimate. Output price matters heavily for long generations, while repeated large prompts can make cache reads valuable. If Fable 5 is under consideration, the separate Fable 5 pricing guide provides a narrower model-level view.
How to estimate Claude API cost
For a basic request, multiply input tokens by the model’s input rate and output tokens by its output rate, then divide each by one million. Add cache, tool, location, batch, or Fast Mode adjustments only when the workload actually uses them.
- Input cost = input tokens ÷ 1,000,000 × input rate.
- Output cost = output tokens ÷ 1,000,000 × output rate.
- Request cost = input cost + output cost + applicable cache and tool charges.
The example below uses one million uncached input tokens and 250,000 output tokens. It is a transparent unit example, not a forecast of a typical user’s monthly bill.
| Model | 1M input | 250K output | Example total |
|---|---|---|---|
| Fable 5 | $10.00 | $12.50 | $22.50 |
| Opus 5 | $5.00 | $6.25 | $11.25 |
| Sonnet 5 | $2.00 | $2.50 | $4.50 |
| Haiku 4.5 | $1.00 | $1.25 | $2.25 |
Batch, caching, location, Fast Mode, and tool costs
A token-only estimate can still be incomplete. Batch processing can lower eligible asynchronous work, while US-only inference, Fast Mode, managed agents, web search, and code execution can add charges or change the effective rate.
| Item | Verified public rate | Planning note |
|---|---|---|
| Batch processing | 50% off | Useful for asynchronous workloads that do not need an immediate response. |
| US-only inference | 1.1x input and output | A location-specific premium; do not apply it to ordinary global processing. |
| Opus 5 Fast Mode | 2x standard price | Up to about 2.5x faster according to the pricing page. |
| Managed Agents runtime | $0.08/session-hour | Add runtime cost to token cost. |
| Web search | $10/1K searches | Input/output tokens used to process results are additional. |
| Code execution | 50 org-hours/day free; then $0.05/hour/container | Model token costs still apply. |
The cache prices in the main model table are for a five-minute TTL. Extended caching has separate documentation and should be priced from the configuration actually used.
Claude application limits vs API limits
Claude application limits are capacity rules, not a guaranteed message count. Anthropic describes a rolling five-hour session window, with weekly limits on paid plans. Conversation length, complexity, model choice, and features change how quickly capacity is consumed.
| Limit system | How it works | What it is not |
|---|---|---|
| Claude app / Claude Code plan allowance | Rolling five-hour session window; paid plans also have weekly limits; capacity depends on model, conversation length, and features | A fixed number of messages |
| API spend limit | Maximum monthly organization spend by usage tier: Start $500, Build $1,000, Scale $200,000; Custom by agreement | A Pro or Max usage window |
| API rate limit | Requests and input/output tokens allowed over time; short bursts can trigger 429 responses | A fixed-time subscription reset |
For API organizations, spend caps and rate limits are separate controls. Public monthly spend caps were $500 for Start, $1,000 for Build, and $200,000 for Scale on the check date; Custom arrangements go through the account team. Rate limits use a token-bucket model, so a short burst can fail even when the monthly spend budget remains available.
GlobalGPT as a separate multi-model subscription
GlobalGPT is relevant when the real job spans several providers rather than only Claude. Its subscription can support a connected workflow across chat, research, writing, coding help, images, video, and audio without maintaining a separate subscription for every tool.
| Plan | Annual billing: displayed monthly equivalent | Monthly promotion on check date | Credits |
|---|---|---|---|
| Basic | $5.80/month | $8.90/month | 144,000/year or 12,000/month |
| Pro | $10.80/month | $15/month | 240,000/year or 20,000/month |
| Unlimited | $25/month | $37/month | 624,000/year or 52,000/month; unlimited use only on selected models |
The monthly figures were promotional values visible on the check date, while the annual figures are displayed monthly equivalents for annual billing. Check the live GlobalGPT pricing page before purchasing. If official Claude controls or direct Anthropic API billing are mandatory, use the official route; if everyday work benefits from several model families, compare the multi-model workflow instead.
Which Claude pricing option should you choose?
Choose from the work backward. The common mistake is buying the highest plan before measuring interruptions, or choosing a low API rate without accounting for output length, tools, and retries.
| Your real job | Best starting route | Why |
|---|---|---|
| Occasional Claude chat | Claude Free | No subscription until usage interruptions matter. |
| Daily individual writing, analysis, or coding | Claude Pro | The first paid tier, with Claude Code included in the shared allowance. |
| Pro limits interrupt billable work | Claude Max 5x or 20x | Pay for more capacity; choose the tier from actual interruption frequency. |
| Managed access for 2–150 people | Team Standard or Premium | Seats, organization controls, and capacity matter more than a personal-plan comparison. |
| Production app or agent | Claude API | Token, cache, tool, spend, and rate-limit economics are the relevant budget. |
| Everyday work across several AI providers | GlobalGPT | A separate multi-model workspace can reduce subscription switching, but does not replace the official API or enterprise controls. |
- Start on Free if Claude is still an experiment.
- Move to Pro when Free limits interrupt regular individual work.
- Move from Pro to Max only when capacity interruptions justify a fivefold or tenfold monthly price jump.
- Use Team or Enterprise when seats, administration, compliance, and organization billing are part of the requirement.
- Use the API for software and agents; monitor tokens, tools, spend limits, and rate limits separately.
- Consider a multi-model subscription when the workflow—not official Claude feature parity—is the value.
If the decision is broader than Claude alone, review the best Claude alternatives before committing to another standalone subscription.
Frequently asked questions
How much is Claude Pro in 2026?
Claude Pro is $20 per month or $200 prepaid annually. Anthropic displays the annual plan as a $17 monthly equivalent. Prices were checked on August 18, 2026 and may exclude applicable taxes.
How much is Claude Max?
Claude Max 5x is $100 per month and Max 20x is $200 per month. Anthropic’s public pages conflict on Max annual billing, so verify the purchase screen instead of assuming an annual price or discount.
Does Claude Pro include API access?
Claude Pro can include Claude Code and Claude app features, but it does not provide general API credits for applications. API use is billed separately by tokens and related services.
Is Claude Code included with Pro and Max?
Yes. Pro and Max can cover Claude Code, but Claude Code shares the plan allowance with Claude on web, desktop, mobile, and supported IDEs. If an ANTHROPIC_API_KEY is configured, API billing takes priority.
How often do Claude usage limits reset?
Claude plan usage uses a rolling five-hour session window. Paid plans also have weekly limits, and Anthropic can apply model, feature, monthly, or other caps. There is no universal fixed message count.
How is the Claude API priced?
The Claude API charges separately for input tokens, output tokens, prompt caching, and some tools. Rates are stated per million tokens, and output tokens usually cost more than input tokens.
Is GlobalGPT cheaper than Claude Max?
GlobalGPT’s displayed subscription entry prices were lower on August 18, 2026, but the products and units are not equivalent. GlobalGPT is a multi-model workspace with platform credits; Claude Max is an official Claude application plan with higher Claude capacity.
