Subscription details · Source-backed

Anthropic API

Focused Claude API for text, code, and agent workloads.

Verification status
Outdated
Data checked
Price locale
English · USD
Published facts
25

Cost

Plans and pricing

All amounts match the page locale; each conversion is disclosed.

paid

Fable 5 input / 1M tokens

$10 / usage unit Tax status not publicly specified

Fable 5 input

  • Input billing for the Fable model.
  • For demanding text and agent workflows.
  • Price and scope are linked in the official source.
Price source ↗

paid

Fable 5 cache write / 1M tokens

$12.50 / usage unit Tax status not publicly specified

Fable 5 cache write

  • Billing for writing into the prompt cache.
  • For long system prompts and stable tool schemas.
  • Price and scope are linked in the official source.
Price source ↗

paid

Fable 5 cache read / 1M tokens

$1 / usage unit Tax status not publicly specified

Fable 5 cache read

  • Reduced input billing on cache hits.
  • Scales to long conversations and agent loops.
  • Price and scope are linked in the official source.
Price source ↗

paid

Fable 5 output / 1M tokens

$50 / usage unit Tax status not publicly specified

Fable 5 output

  • Output billing for the Fable model.
  • For more extensive model responses.
  • Price and scope are linked in the official source.
Price source ↗

paid

Opus 4.8 input / 1M tokens

$5 / usage unit Tax status not publicly specified

Opus 4.8 input

  • Input billing for a high-capability Claude model.
  • Suited to complex reasoning and text tasks.
  • Price and scope are linked in the official source.
Price source ↗

paid

Opus 4.8 cache write / 1M tokens

$6.25 / usage unit Tax status not publicly specified

Opus 4.8 cache write

  • Billing for writing into the prompt cache.
  • Lowers cost for reusable contexts.
  • Price and scope are linked in the official source.
Price source ↗

paid

Opus 4.8 cache read / 1M tokens

$0.50 / usage unit Tax status not publicly specified

Opus 4.8 cache read

  • Reduced input billing on cache hits.
  • For enterprise and agent workloads with long contexts.
  • Price and scope are linked in the official source.
Price source ↗

paid

Opus 4.8 output / 1M tokens

$25 / usage unit Tax status not publicly specified

Opus 4.8 output

  • Output billing for Opus responses.
  • For detailed results in complex workflows.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 5 input / 1M tokens

$2 / usage unit Tax status not publicly specified

Sonnet 5 input

  • Introductory input billing for Sonnet 5.
  • For high-performance coding and agent tasks.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 5 cache write / 1M tokens

$2.50 / usage unit Tax status not publicly specified

Sonnet 5 cache write

  • Billing for writes to the Sonnet 5 prompt cache.
  • For reusable long contexts and tool definitions.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 5 cache read / 1M tokens

$0.20 / usage unit Tax status not publicly specified

Sonnet 5 cache read

  • Reduced input billing on cache hits.
  • For recurring agent and coding contexts.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 5 output / 1M tokens

$10 / usage unit Tax status not publicly specified

Sonnet 5 output

  • Introductory billing for Sonnet 5 output.
  • For model responses in coding and agent workflows.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 4.6 input / 1M tokens

$3 / usage unit Tax status not publicly specified

Sonnet 4.6 input

  • Input billing for Sonnet tasks.
  • A balanced option for text and coding work.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 4.6 cache write / 1M tokens

$3.75 / usage unit Tax status not publicly specified

Sonnet 4.6 cache write

  • Billing for writing into the prompt cache.
  • Saves cost in typical production contexts.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 4.6 cache read / 1M tokens

$0.30 / usage unit Tax status not publicly specified

Sonnet 4.6 cache read

  • Reduced input billing on cache hits.
  • For recurring system prompts and tool definitions.
  • Price and scope are linked in the official source.
Price source ↗

paid

Sonnet 4.6 output / 1M tokens

$15 / usage unit Tax status not publicly specified

Sonnet 4.6 output

  • Output billing for Sonnet responses.
  • For responses in production applications.
  • Price and scope are linked in the official source.
Price source ↗

paid

Haiku 4.5 input / 1M tokens

$1 / usage unit Tax status not publicly specified

Haiku 4.5 input

  • Input billing for fast Claude workflows.
  • For frequent, lighter requests.
  • Price and scope are linked in the official source.
Price source ↗

paid

Haiku 4.5 cache write / 1M tokens

$1.25 / usage unit Tax status not publicly specified

Haiku 4.5 cache write

  • Billing for writing into the prompt cache.
  • For high-throughput classifiers and small agents.
  • Price and scope are linked in the official source.
Price source ↗

paid

Haiku 4.5 cache read / 1M tokens

$0.10 / usage unit Tax status not publicly specified

Haiku 4.5 cache read

  • Reduced input billing on cache hits.
  • Very low read rates for high-frequency requests.
  • Price and scope are linked in the official source.
Price source ↗

paid

Haiku 4.5 output / 1M tokens

$5 / usage unit Tax status not publicly specified

Haiku 4.5 output

  • Output billing for Haiku responses.
  • For fast responses at high throughput.
  • Price and scope are linked in the official source.
Price source ↗

Editorial

Pros and cons

Editorial assessment, separate from provider claims.

Pros

  • Strong models for coding and long documents.
  • Prompt caching and batch processing can reduce costs.
  • Clear developer documentation and tool-use capabilities.

Cons

  • No native image generation and a smaller media portfolio.
  • Rate limits and costs scale with model and usage tier.

Provider claims

What the provider says is included

Plan statements from the official feature overview, not our assessment. View original ↗

API mode

FeatureFable 5 input / 1M tokensFable 5 cache write / 1M tokensFable 5 cache read / 1M tokensFable 5 output / 1M tokensOpus 4.8 input / 1M tokensOpus 4.8 cache write / 1M tokensOpus 4.8 cache read / 1M tokensOpus 4.8 output / 1M tokensSonnet 5 input / 1M tokensSonnet 5 cache write / 1M tokensSonnet 5 cache read / 1M tokensSonnet 5 output / 1M tokensSonnet 4.6 input / 1M tokensSonnet 4.6 cache write / 1M tokensSonnet 4.6 cache read / 1M tokensSonnet 4.6 output / 1M tokensHaiku 4.5 input / 1M tokensHaiku 4.5 cache write / 1M tokensHaiku 4.5 cache read / 1M tokensHaiku 4.5 output / 1M tokens
Usage-based billingIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncludedIncluded
Pricing modelOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed planOfficially listed plan
Verification sourceOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overviewOfficial provider overview

Provider facts

Features, limits, and restrictions

A concise view of published statements. Full evidence follows below.

Included and promised

Not publicly specified

Usage limits

  • Usage limit: Requests, tokens, and spend are limited by usage tier.

Resets, privacy, and availability

  • Reset: Exact account limits depend on tier and model.

Platforms and ecosystems

  • Platform: Claude models are available through the Messages API and SDKs.
  • Ecosystem: The API supports prompt caching, Batch, and tools.

Restrictions and geo-blocking

  • Restriction: The API is metered and billed in USD.

Evidence

Sources and checked dates

Summaries instead of copied provider text; every entry links to the original page.

Original source

Anthropic API pricing

Open source
  • Official fact

    Official price for Fable 5 input / 1M tokens (usage).

  • Official fact

    Official price for Fable 5 cache write / 1M tokens (usage).

  • Official fact

    Official price for Fable 5 cache read / 1M tokens (usage).

  • Official fact

    Official price for Fable 5 output / 1M tokens (usage).

  • Official fact

    Official price for Opus 4.8 input / 1M tokens (usage).

  • Official fact

    Official price for Opus 4.8 cache write / 1M tokens (usage).

  • Official fact

    Official price for Opus 4.8 cache read / 1M tokens (usage).

  • Official fact

    Official price for Opus 4.8 output / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 5 input / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 5 cache write / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 5 cache read / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 5 output / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 4.6 input / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 4.6 cache write / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 4.6 cache read / 1M tokens (usage).

  • Official fact

    Official price for Sonnet 4.6 output / 1M tokens (usage).

  • Official fact

    Official price for Haiku 4.5 input / 1M tokens (usage).

  • Official fact

    Official price for Haiku 4.5 cache write / 1M tokens (usage).

  • Official fact

    Official price for Haiku 4.5 cache read / 1M tokens (usage).

  • Official fact

    Official price for Haiku 4.5 output / 1M tokens (usage).

Original source

Anthropic API Docs

Open source
  • Official fact

    Claude models are available through the Messages API and SDKs.

Original source

Anthropic Pricing

Open source
  • Official fact

    The API supports prompt caching, Batch, and tools.

  • Official fact

    The API is metered and billed in USD.

Original source

Anthropic Rate Limits

Open source
  • Official fact

    Requests, tokens, and spend are limited by usage tier.

  • Not publicly specified

    Exact account limits depend on tier and model.

Transparency

Confirmed changes

No confirmed change has been published since automated tracking began.