🏔️ Anthropic Claude API Pricing (Per 1 Million Tokens)

Anthropic categorizes its models into a "family" based on a balance of intelligence, speed, and cost: Opus (highest intelligence), Sonnet (best balance), and Haiku (fastest and most cost-effective).

Model Category

Model Name

Input Price (per 1M Tokens)

Output Price (per 1M Tokens)

Use Case & Speed

Opus (Highest Tier)

Claude 4.1 / 4.0 Opus

$15.00

$75.00

Highly complex tasks, advanced reasoning, and strategic analysis.

Sonnet (Balanced Tier)

Claude 4.5 / 4.0 Sonnet

$3.00

$15.00

High-throughput tasks, data processing, and code generation.

Haiku (Fastest Tier)

Claude 4.5 / 4.0 Haiku

$1.00

$5.00

Simple, high-speed tasks, like quick support and content moderation.

Haiku (Older Tier)

Claude 3.5 Haiku

$0.80

$4.00

Even faster, for simple tasks where speed is critical.

Note on Context Window: A major selling point for Claude models is their large context window (the amount of text the model can 'remember' at once). Most Claude models above support up to 200,000 tokens by default, which is significantly larger than the base offering of many competitors.


💰 Anthropic API Pricing Features

Anthropic also offers different mechanisms that can greatly affect your total billing:

Feature

Cost Impact

Model Tier Example (Sonnet)

Prompt Caching (Read)

Low cost to reuse a cached prompt.

$\approx \$0.30$ per 1M tokens

Prompt Caching (Write)

Low cost to write a new prompt to the cache.

$\approx \$3.75$ per 1M tokens

Batch Processing

Up to 50% cost savings for large, non-urgent jobs.

Significantly reduces base Input and Output costs.



 

Token-to-Word Reminder

As a reminder, the conversion rule of thumb remains the same for the Claude models:

1,000,000 tokens (for English) $\approx 750,000$ words.


Disclaimer: Prices for the older Claude 3 series (Opus 3, Sonnet 3, Haiku 3) are still available but generally higher than the newer 4.x/4.x models (as the newer models are often more efficient). Always check the official Anthropic pricing page for the most accurate and up-to-date figures.