Claude API pricing

Anthropic sells four current Claude models through the API, and the price gap between them is 100x on input. On October 21, 2026, Anthropic listed Claude Haiku 5.5 at $0.10 input and $0.50 output per million tokens, Claude Sonnet 5.5 at $2 and $10, Claude Opus 5.5 at $4 and $20, and Claude Fable 5.1 at $10 and $50.

Claude Haiku 5.5 is the only current Claude model with a long-prompt price: prompts over 100,000 tokens cost $0.50 input and $2.50 output per million. Claude Sonnet 5.5, Opus 5.5 and Fable 5.1 charge the same rate across their full 1 million token context window.

Claude models from version 4.7 on use a new tokenizer that produces about 30% more tokens for the same text, according to Anthropic. A per-token price that looks equal to an older Claude model therefore costs more for the same document.

Claude prompt caching charges more once to store a prompt prefix and far less every time it is read again. A cache write costs 1.25x the input price for a 5 minute cache and 2x for a 1 hour cache. A cache read costs 0.1x the input price on most models, 0.05x on Claude Opus 5.5 and Sonnet 5.5, and 0.025x on Claude Fable 5.1.

Claude prompt caching pays off from the second read of the same prefix. The cache lifetime resets for free each time the prefix is used, and the minimum cacheable prompt is 512 tokens on the current models.

The Claude Message Batches API halves both input and output prices, so Claude Sonnet 5.5 costs $1 input and $5 output per million tokens in a batch. Anthropic says most batches finish in under an hour, and a batch expires after 24 hours.

One Claude batch holds up to 100,000 requests or 256 MB, and results stay available for 29 days. Batch discounts stack with prompt caching. Fast mode is not available in batches.

Claude fast mode on Opus 5.5 costs twice the standard rate: $8 input and $40 output per million tokens. Running inference only in the US (inference_geo set to "us") adds 10% to every token category on Claude models from 4.6 on; the default global routing has no surcharge.

Claude tools add their own fees on top of tokens. Web search costs $10 per 1,000 searches, web fetch is free apart from tokens, and code execution is free with web search or web fetch, otherwise 1,550 free container hours a month per organization and then $0.05 per hour.

A Claude subscription and the Claude API are billed separately: Anthropic states that a paid Claude plan "enhances your chat experience, and the Claude API is billed separately". Claude Pro costs $20 a month ($17 a month billed yearly), Claude Max costs $100 or $200 a month, and Team starts at $25 per seat a month.

Claude Code is included in Claude Pro and Max, and usage is shared with Claude on the web, desktop and mobile within a 5 hour window plus a weekly limit. Setting an ANTHROPIC_API_KEY switches Claude Code to API billing at the rates above. Max and Team plans also include monthly Claude API credits, which cannot be spent in Claude Code or the Claude apps.

For a rough break even, $20 buys 2 million Claude Sonnet 5.5 output tokens on the API. A developer who stays within Claude Pro limits every week gets more work per dollar from the subscription; an application serving other people needs the API, because a subscription cannot be called from your own product.

The Claude API has no free tier on Anthropic's pricing pages, checked on October 21, 2026: every token is paid from the Console account. The free Claude plan covers the chat apps only and does not include API access or Claude Code.

A Claude API key is created in the Claude Console under Settings, API keys (platform.claude.com/settings/keys). The key starts with sk-ant- and is shown once.

Claude API rate limits grow with spend tiers. The Start tier caps spend at $500 a month with 1,000 requests per minute on the 5.5 models, Build allows $1,000 a month and 5,000 requests per minute, and Scale allows $200,000 a month and 10,000 requests per minute. Anthropic moves accounts up automatically based on usage history, and cache reads do not count toward the input token limit.

• June 30, 2026: Claude Sonnet 5 launched at $2 input and $10 output; on August 10 Anthropic kept that price instead of the announced rise to $3 and $15.
• September 1, 2026: Claude Fable 5.1 cut the cache read price from $1 to $0.25 per million tokens.
• September 22, 2026: Claude Opus 5.5 launched at $4 and $20, 20% below Claude Opus 5 at $5 and $25.
• October 7, 2026: Claude Haiku 5.5 launched at $0.10 and $0.50, 90% below Claude Haiku 4.5 at $1 and $5.

Anthropic retires older Claude models on published dates. Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) is deprecated and retires on November 30, 2026, with Claude Sonnet 5.5 as the named replacement. Claude Opus 4.1, Opus 4, Sonnet 4, Haiku 3, Haiku 3.5 and Sonnet 3.7 were retired between February and August 2026.