This site is not affiliated with, endorsed by, or connected to Anthropic, PBC. Claude and the Claude logo are trademarks of Anthropic. All pricing shown is sourced from public Anthropic documentation. Verify current pricing at claude.com/pricing.
Updated April 2026

Claude Token Costs 2026: What Tokens Are, What They Cost, and How to Calculate Your Bill

Tokens are the unit of measurement for Claude API billing. Subscriptions (Pro, Max, Team) do not use token billing. Here is everything you need to understand costs.

What Is a Token?

A token is roughly 4 characters or 0.75 words in English text. Claude tokenizes text differently to how humans read words - common words are single tokens, rare or long words may be multiple tokens.

Content TypeApproximate SizeApproximate Tokens
Tweet / short message280 characters~70 tokens
Paragraph of text100 words~130 tokens
Standard essay800 words~1,060 tokens
Short story5,000 words~6,700 tokens
Academic paper8,000 words~10,700 tokens
Code file (500 lines)~15,000 characters~3,500 tokens
Large codebase file (2000 lines)~60,000 characters~14,000 tokens
Novel chapter20,000 words~26,700 tokens

API Token Pricing by Model

ModelInput (per 1M tokens)Output (per 1M tokens)Batch InputBatch OutputContext Window
Claude Opus 4.6$5.00$25.00$2.50$12.501M (API)
Claude Sonnet 4.6$3.00$15.00$1.50$7.50200K
Claude Haiku 4.5$1.00$5.00$0.50$2.50200K
Claude Haiku 3.5$0.80$4.00$0.40$2.00200K

Token Cost Calculator

Estimate your Claude API costs based on word count. This is for API billing only - subscriptions (Pro, Max, Team) do not charge by token.

Quick presets:

Input tokens

665

Output tokens

266

Cost per request

$0.0060

Monthly total

$5.99

At this usage level, a Claude Pro subscription at $20/month would give you approximately the same usage plus Claude Code and Cowork - potentially better value for individual use.

Subscriptions Do NOT Bill by Token

Common confusion clarified

Claude Pro, Max, Team, and Enterprise plans do NOT use token-based billing. The question "how many tokens do I get with Claude Pro?" does not have a direct answer because tokens are not the unit of account.

Subscriptions use a usage-window model. Anthropic measures compute usage (which is related to but not identical to token count) in a rolling 5-hour window. When you hit your window limit, you wait for it to reset - there is no per-token charge.

Frequently Asked Questions

Do Claude subscriptions charge by token?
No. Claude Pro, Max, Team, and Enterprise subscriptions do not bill by token. Subscriptions use a usage-window model based on compute time and the complexity of your requests. You never see a token count or a per-token charge on a subscription plan. Token-based billing only applies to the Anthropic API (accessed via API key through the Anthropic Console).
How do I count tokens before sending a request?
Anthropic provides a token counting API endpoint that returns the exact token count for a given message before you send it. You can also use rough estimates: approximately 750 words per 1,000 tokens, or 4 characters per token. The Anthropic Python and TypeScript SDKs include tokenization utilities for offline estimation.
What is the context window and does it affect cost?
The context window is the maximum amount of text (measured in tokens) that Claude can process in a single request - including both your input and Claude's output. Larger context windows cost more because they require more memory and compute. Claude Sonnet 4.6 has a 200K token context window. At subscription level, you pay a flat rate and the context window is included. At API level, you pay per token for both input and output.
Does Claude count output tokens separately?
Yes. In the API, input tokens (what you send) and output tokens (what Claude generates) are priced separately. Output tokens are typically priced higher because they require more compute to generate. Output tokens are a subset of the context window used for each response. In subscriptions, this distinction does not affect your billing.
How does the Batch API halve costs?
The Batch API processes requests asynchronously with results delivered within 24 hours. Because the compute is scheduled flexibly rather than in real-time, Anthropic can run it at 50% of the standard API price. All models support batch processing. The Batch API is accessible via the same API key - you simply use a different endpoint and accept the latency trade-off.

Related Pages