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 API Pricing 2026: Models, Token Costs, and When to Use API vs Subscription

The Claude API is billed per token - separately from any subscription. Here are the current rates, discount options, and the break-even math for subscription vs API.

Deep API pricing: This page gives an overview of API pricing and the subscription vs API decision. For model-by-model token costs, context window details, prompt caching implementation, and full API rate tables, see claudeapipricing.com.

Claude API Model Pricing (per 1M tokens)

ModelInputOutputBatch InputBatch OutputContext
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

Batch API prices shown are 50% of standard rates. Prices sourced from Anthropic public documentation, April 2026.

API Cost Reduction Options

Batch API

50% off

Submit requests in batches for non-real-time processing. Results within 24 hours. Cuts all model rates in half. Best for pipelines and bulk jobs.

Learn more

Prompt Caching

90% off cached reads

Cache repeated system prompts. First cache write costs 25% more. Every subsequent cached read costs 90% less. Essential for apps with large shared context.

Learn more

Model Selection

Up to 85% off

Haiku 3.5 at $0.80/$4 vs Opus 4.6 at $5/$25 - choosing the right model for the task is the single biggest cost lever available.

Learn more

Subscription vs API: The Break-Even Math

The question is simple: would your usage on the API cost more or less than a subscription?

Daily Usage PatternAPI Cost (Sonnet 4.6)Best Option
5 msgs/day (avg 500 tokens each)~$3/monthFree tier or API
20 msgs/day (avg 1K tokens each)~$12/monthPro at $20/month (includes Claude Code)
50 msgs/day (avg 1.5K tokens each)~$45/monthPro (API is more expensive)
100 msgs/day + Claude Code (2h)~$120-180/monthMax 5x at $100/month
200+ msgs/day + full-day Claude Code~$350-600/monthMax 20x at $200/month

Estimates based on Sonnet 4.6 standard rates. Actual costs vary by message length and model choice.

When API Is the Right Choice

  • Building applications: Subscriptions are for human users on claude.ai. APIs are for building products and automations.
  • Fluctuating usage: API scales to zero. If you only need Claude occasionally, API billing avoids paying for idle subscription months.
  • Per-request transparency: For agencies billing clients per task, API gives exact cost attribution that subscriptions cannot.
  • 1M-token context: Claude Opus 4.6 via API supports a 1M token context window. Subscriptions cap at 200K (500K at Enterprise).
  • Batch processing: Overnight batch jobs at 50% discount are only available via API, not subscriptions.

Frequently Asked Questions

Can I use Claude API with a Pro subscription?
No. Claude subscriptions (Pro, Max, Team, Enterprise) and API access are completely separate billing systems. Subscriptions give you access to claude.ai web, desktop, and mobile apps. API access is a separate Anthropic Console account billed per token. You can have both - many developers use a subscription for personal use and a separate API account for their applications.
Is there a free API tier?
New Anthropic API accounts may receive a small initial credit for testing (historically $5, but this changes). After that, API usage is billed per token with no free allocation. Claude.ai's free plan is separate from API access and does not give you API tokens. Check the Anthropic Console for current new-account credit offers.
When is the Batch API cheaper than real-time?
The Batch API offers a 50% discount on all models for non-real-time workloads. This means Sonnet 4.6 drops from $3/$15 to $1.50/$7.50 per 1M tokens. Use the Batch API when your use case allows for results within 24 hours rather than immediate responses - examples include document processing pipelines, evaluation runs, bulk classification, and overnight analysis jobs.
What is the difference between input and output token pricing?
Input tokens are the tokens you send to Claude (your prompt, system prompt, context). Output tokens are the tokens Claude generates in response. Output tokens are priced higher because they require more compute. For most workloads, outputs are a fraction of inputs by token count, so the effective cost is often closer to the input rate. Claude Code sessions can have high output ratios (Claude writes a lot of code).

Related Pages