Claude API pricing: every model, prompt caching, batch and tools

Claude API prices per million tokens, input and output: Claude Opus 5.5 costs $4 and $20, Sonnet 5.5 $2 and $10, Haiku 5.5 $0.10 and $0.50, and Fable 5.1 $10 and $50. Reading from the prompt cache costs a small fraction of the input price (5% on Opus 5.5 and Sonnet 5.5), the Batch API halves everything, and the current models carry the full 1 million token context window at the standard price, except Haiku 5.5, which costs five times as much above 100,000 tokens. Web search is $10 per 1,000 searches, US-only inference adds 10%, and new accounts start with a small amount of free credit.

Key takeaways

• Flagship: Opus 5.5 at $4 and $20, the same price Cursor and GitHub Copilot charge for it.
• Everyday: Sonnet 5.5 at $2 and $10. Cheapest: Haiku 5.5 at $0.10 and $0.50.
• Top end: Fable 5.1 at $10 and $50.
• Caching: writes cost 1.25 times input for five minutes or 2 times for an hour; reads cost 0.025 to 0.1 times.
• Batch: 50% off input and output for jobs that can wait up to a day.
• Long context: no surcharge on Opus 5.5, Sonnet 5.5 or Fable 5.1.

How much does the Claude API cost?

From Anthropic's pricing docs (Anthropic), per million tokens:

• Claude Fable 5.1: $10 input, $50 output. Cache writes $12.50 (five minutes) or $20 (one hour), cache reads $0.25.
• Claude Opus 5.5: $4 input, $20 output. Cache writes $5 or $8, reads $0.20.
• Claude Sonnet 5.5: $2 input, $10 output. Cache writes $2.50 or $4, reads $0.10.
• Claude Haiku 5.5: $0.10 input, $0.50 output on prompts up to 100,000 tokens; $0.50 and $2.50 above that. Cache reads $0.01.

Older models still on sale cost more than their replacements:

• Opus 5 and Opus 4.5 to 4.8: $5 and $25.
• Sonnet 5: $2 and $10; the planned rise to $3 and $15 "will not occur."
• Sonnet 4 to 4.6: $3 and $15.
• Haiku 4.5: $1 and $5.
• Fable 5: $10 and $50.

Opus 4 and 4.1, at $15 and $75, are retired on Anthropic's own API. Anthropic notes that models from 4.7 onward use "approximately 30% more tokens for the same text", so a per-token price comparison with older models understates the difference.

How much does prompt caching save?

A prompt that starts with the same long prefix every time (a system prompt, a codebase, a document) can be cached:

• Writing to the cache costs 1.25 times the input price for a five-minute cache, or 2 times for one hour.
• Reading costs 0.1 times the input price on most models, 0.05 times on Opus 5.5 and Sonnet 5.5, and 0.025 times on Fable 5.1.

On Sonnet 5.5, a 100,000-token prefix costs $0.20 a call uncached. Cached, the first call writes it for $0.25 and each call after that reads it for $0.01. Ten calls inside five minutes cost about $0.34 instead of $2. Coding agents depend on this: every turn resends the whole conversation, and caching is what keeps long sessions affordable (how to cut Claude Code token use).

What does the Batch API cost?

Requests sent through the Message Batches API cost 50% less, input and output: Opus 5.5 becomes $2 and $10, Sonnet 5.5 $1 and $5. Results come back asynchronously, within 24 hours. It suits evaluations, classification and bulk summaries, not anything a person is waiting on. Batch discounts stack with caching.

Does the Claude API charge more for long context?

Not on current models. Anthropic says "Claude 4.6 and later models (except Claude Haiku 5.5)" include the full 1 million token context window at standard pricing, so a 500,000-token request on Sonnet 5.5 costs exactly 500,000 times the per-token rate. Haiku 5.5 is the exception: prompts over 100,000 tokens cost five times as much, $0.50 input and $2.50 output.

What else costs extra?

• Fast mode: Opus 5.5 at $8 and $40, double the standard price, for faster output. Not available with Batch.
• US-only inference: setting inference to the US multiplies every token price by 1.1, on 4.6 and later models.
• Web search: $10 per 1,000 searches, plus the tokens the results add. Web fetch has "no additional cost" beyond tokens.
• Code execution: free when used with web search or web fetch; otherwise $0.05 an hour per container, after a free allowance.
• Managed Agents: $0.08 per session-hour, plus tokens.

What are the Claude API usage tiers?

New organizations get "a small amount of free credits" and may start in an evaluation tier. After that, tiers set a monthly spend cap (Anthropic):

• Start: up to $500 a month.
• Build: up to $1,000.
• Scale: up to $200,000.
• Custom: no cap, by arrangement.

At Start, Opus 5.5, Sonnet 5.5 and Haiku 5.5 each allow 1,000 requests, 2 million input tokens and 400,000 output tokens a minute.

API or a Claude subscription?

For one person using Claude Code every day, a subscription is usually cheaper: Anthropic puts the average Claude Code user at about $13 per active day at API prices, while Claude Pro costs $20 a month and Max $100 or $200 (Claude Code pricing). The API suits apps, automations and occasional heavy jobs.

Since October 7, 2026, subscriptions also come with API credit: $100 a month on Max 5x, $200 on Max 20x, and on Team $20 per standard seat and $100 per premium seat, up to $500. The credits cover the API, Managed Agents and the Agent SDK, not Claude Code, and expire each billing cycle (Claude Team plan pricing). Claude Enterprise bills all its usage at these API rates (Claude Enterprise pricing).

Claude API vs OpenAI API prices

• Top models: Opus 5.5 at $4 and $20 against GPT-6 Astra at $10 and $50.
• Mid-range: Sonnet 5.5 and GPT-6.1 Sol both at $2 and $10.
• Small: Haiku 5.5 and GPT-6 Luna both at $0.10 and $0.50.

OpenAI charges extra above 272,000 input tokens; Anthropic doesn't on its current models apart from Haiku 5.5. Full list in OpenAI API pricing, and Fable's place in the lineup in Claude Fable pricing.

Claude without an API key

Poly (usepoly.co) runs Claude, GPT, Grok and Muse for a whole team in one shared room, billed as credits per person, so nobody needs an API key or a usage tier. Sonnet 5.5 and Haiku 5.5 are free to start. What is Poly?