The Complete Guide to Claude Pricing: Plans, API Costs and How to Budget
How Anthropic prices Claude in 2026, from subscription plans to API tokens, and how founders and app builders can estimate and control their costs.
Two Ways to Pay for Claude
Anthropic charges for Claude in two separate ways, and the first decision is which one applies to you:
-
✓
Subscription plans for people who use Claude directly through claude.ai and the Claude apps. You pay a fixed monthly fee per person.
-
✓
API pricing for developers who build Claude into their own products. You pay for what you use, measured in tokens.
They are billed separately: a Claude Pro subscription does not include API usage, and API credits do not unlock claude.ai features. Prices and plan details change, so always confirm current figures on Anthropic’s own pricing pages before you budget.
Claude Plans for Individuals and Teams
Free
Limited access to Claude for trying it out, with usage caps.
Pro
A monthly plan for individuals, with higher usage limits and access to more capable models and features.
Max
Higher-priced tiers for heavy individual users who need substantially more usage than Pro.
Team
Per-seat pricing for organisations, with shared workspaces and central administration.
Enterprise
Custom contracts with security, compliance and administration features for larger organisations.
API
Pay-as-you-go access for developers, billed per token by model.
How Claude API Pricing Works
API usage is priced per million tokens, where a token is roughly three quarters of an English word. Three factors decide what you pay:
The Model Tier
Claude models come in three tiers. Haiku is the fastest and cheapest, suited to classification, routing and short summaries. Sonnet balances quality and cost, and suits most product features. Opus is the most capable and most expensive, for complex reasoning where accuracy matters most. Anthropic releases new versions regularly, so check which models are current before you build.
Input vs Output Tokens
You pay separately for the text you send (input) and the text Claude writes (output). Output tokens cost several times more than input tokens, so long answers cost more than long prompts.
Discounts for How You Call It
Prompt caching lets you reuse large, repeated context such as instructions or reference documents at a much lower cost. Batch processing, for jobs that do not need an instant answer, is offered at a discount. Both can cut costs substantially for the right workloads.
For current per-token rates by model, see Anthropic’s API pricing page. Rates change as new models launch, so avoid hard-coding old prices into your budget.
How to Estimate Your Monthly Claude API Cost
You can build a reasonable estimate before writing any code. Multiply your expected calls per month by the tokens per call, then by the price per token for the model you plan to use.
A worked example with illustrative rates (not current prices): suppose a model costs $3 per million input tokens and $15 per million output tokens. A call with 500 input tokens and 300 output tokens then costs about $0.006. At 10,000 calls a month that is about $60, and at 100,000 calls about $600. Swap in the real rates from Anthropic’s pricing page for the model you choose.
The same method shows where costs can run away: sending whole documents on every call, using the most expensive tier for simple tasks, or generating long answers nobody reads.
Route Each Task to the Right Tier
Use Haiku for quick classification and routing, Sonnet for most content and analysis, and Opus only where its extra capability pays for itself. Matching the model to the task is usually the single biggest cost lever.
Keeping Claude Costs Under Control in a Bubble.io App
Claude connects to Bubble.io through the API Connector, so your workflows can call it and store the results in your database. A few habits keep costs predictable:
-
✓
Decide what each AI feature does and how often it runs before choosing a model
-
✓
Send only the relevant part of a document, not the whole file, on each call
-
✓
Use prompt caching for long, repeated instructions or reference material
-
✓
Store outputs in your Bubble database instead of regenerating identical answers
-
✓
Set spend limits in Anthropic’s console, especially during testing
-
✓
Log each call so you can see which features drive your costs, and remember Bubble’s own workload units are a separate cost
For a step-by-step setup, see our guide to connecting Claude to Bubble.io.
Frequently Asked Questions
Is Claude free to use?
There is a free plan on claude.ai with usage limits. Paid plans such as Pro and Max raise those limits. Building Claude into your own app requires an API account with usage-based billing.
What is the cheapest Claude model on the API?
The Haiku tier is Anthropic’s fastest and lowest-cost model family, suited to high-volume, simpler tasks. Check Anthropic’s pricing page for the current Haiku version and its rates.
Does a Claude Pro subscription include API access?
No. Subscriptions cover use of claude.ai and the Claude apps. API usage is billed separately through Anthropic’s developer platform.
How can I reduce my Claude API bill?
Match each task to the cheapest model that does it well, send only the context you need, use prompt caching for repeated instructions, batch jobs that are not urgent, and store results rather than regenerating them.
Can I use Claude in a Bubble.io app?
Yes. Bubble connects to Claude through its API Connector, so you can call Claude from workflows and save the results in your database.
Ready to Build an AI-Powered App on Bubble.io?
SA Solutions is a Bubble.io product studio led by Athar Ahmad, a certified Bubble.io developer. We build AI features with Claude and other models into Bubble apps, with a cost model planned from the start. Book a free Idea Audit call to talk through your idea.