API pricing

Anthropic Pricing

Anthropic charges from $1.25 per 1 million output tokens, depending on the model. Claude 3 Haiku is the cheapest at $1.25/1M output and $0.250/1M input; its current flagship Claude Fable 5 costs $50.00/1M output. The largest context window is 1.0M tokens (Claude 4 Sonnet (2025-05-14)). Prices are USD, current as of August 11, 2026.

Anthropic prices Claude in three tiers — Haiku for high-volume classification and extraction, Sonnet for general production work, and Opus for the hardest reasoning. Output tokens cost roughly 5× input across the lineup, so response length drives your bill far more than prompt size.

Last updated · synced weekly from the upstream model catalog

Cheapest input

$0.250/1M

Claude 3 Haiku

Cheapest output

$1.25/1M

Claude 3 Haiku

Longest context

1.0M

Claude 4 Sonnet (2025-05-14)

Models priced

22

Avg $26.11/1M output

Pricing by model

All prices in USD per 1 million tokens. All 22 priced models, cheapest first. The batch column is the asynchronous batch-API rate, 50% below the standard output rate on the 19 models that publish one.

ModelInput / 1MOutput / 1MBatch out / 1MContext
Claude 3 Haiku$0.250$1.25200K
Claude 3.5 Haiku$0.800$4.00200K
Claude Haiku 4.5$1.00$5.00$2.5050%200K
Claude Haiku 4.5 (Regional)$1.10$5.50200K
Claude Sonnet 5
VisionReasoningToolsCache
$2.00$10.00$5.0050%1.0M
Claude 3.5 Sonnet$3.00$15.00200K
Claude 3.7 Sonnet$3.00$15.00200K
Claude 3.7 Sonnet (thinking)$3.00$15.00200K
Claude Sonnet 4$3.00$15.001.0M
Claude Sonnet 4.5$3.00$15.00$7.5050%1.0M
Claude Sonnet 4.6$3.00$15.00$7.5050%1.0M
Claude Sonnet 4.5 (Regional)$3.30$16.501.0M
Claude Sonnet 4.5 (Long Context)$6.00$22.501.0M
Claude Opus 4.5$5.00$25.00$12.5050%200K
Claude Opus 4.6$5.00$25.00$12.5050%200K
Claude Opus 4.7$5.00$25.00$12.5050%200K
Claude Opus 4.8
VisionReasoningToolsCache
$5.00$25.00$12.5050%1.0M
Claude Opus 5
VisionReasoningToolsCache
$5.00$25.00$12.5050%1.0M
Claude Fable 5
VisionReasoningToolsCache
$10.00$50.00$25.0050%1.0M
Claude 3 Opus (2024-02-29)$15.00$75.00200K
Claude Opus 4$15.00$75.00200K
Claude Opus 4.1$15.00$75.00$37.5050%200K

A dash in the batch column means our catalog carries no published batch rate for that model — not that the model has no batch tier. Batch coverage in the upstream pricing data is uneven.

Frequently asked questions

How much does Anthropic cost per 1M tokens?

Anthropic pricing starts at $1.25 per 1 million output tokens (Claude 3 Haiku) across 22 models, with its current flagship Claude Fable 5 at $50.00. Older premium models in the catalog list higher. Rates current as of August 11, 2026.

What is the cheapest Anthropic model?

Claude 3 Haiku is the cheapest Anthropic model on both axes — $0.250 per 1M input tokens and $1.25 per 1M output. It is the right default for high-volume, low-complexity work such as classification, extraction and routing.

What is the largest Anthropic context window?

Claude 4 Sonnet (2025-05-14) has the largest context window in the Anthropic catalog at 1.0M input tokens, with up to 64K output tokens per response.

Does Anthropic support prompt caching?

Yes. Claude Sonnet 5 reads cached input at $0.200 per 1M tokens, against $2.00 per 1M uncached. Caching pays off when a long system prompt or document is reused across many requests.

Does Anthropic offer batch pricing?

Yes — 19 Anthropic models in this catalog publish a discounted rate for asynchronous batch jobs, at 50% off output tokens. Claude Opus 5 bills $12.50 per 1M output in batch against $25.00 synchronously. Batch trades real-time responses for the lower rate, so it suits backfills, evaluations and bulk enrichment rather than user-facing calls.

Which Anthropic models support reasoning or vision?

The Anthropic catalog includes 4 reasoning models, 4 vision models, 4 with tool calling. Reasoning models bill their internal thinking as output tokens, so they cost more per visible response than the headline rate suggests.

How do I calculate my actual Anthropic bill?

Multiply your input tokens by the input rate and your output tokens by the output rate, both per million, then add retries and any conversation history resent on each turn. Calcaas does this against your real usage pattern and shows the margin left at your price point.

Head-to-head comparisons

Other providers

What will this actually cost you?

Per-token rates only tell you half the story. Calcaas multiplies them by your real usage — prompt size, response length, retries and conversation history — then shows the margin left at your price point.

Open the free cost calculator