Claude API Pricing (November 2026)
Claude API pricing starts at $0.10 per million input tokens on Haiku 5.5. Sonnet 5.5 costs $2 input and $10 output, Opus 5.5 costs $4 input and $20 output, and the top tier, Fable 5.2, costs $8 input and $40 output. Cached input and the Batch API cost less. Each rate below was read on Anthropic's pricing page on the date in its row.
| Model | Input /M | Cache read /M | Output /M | Batch in /M | Batch out /M | Context | Status | Checked |
|---|---|---|---|---|---|---|---|---|
Claude Opus 5.5 claude-opus-5-5 | $4.00 | $0.20−95% | $20.00 | $2.00 | $10.00 | 1M | Current | 2026-10-19 |
Claude Sonnet 5.5 claude-sonnet-5-5 | $2.00 | $0.10−95% | $10.00 | $1.00 | $5.00 | 1M | Current | 2026-10-09 |
Claude Haiku 5.5 claude-haiku-5-5 | $0.10$0.50 over 100K | $0.01−90% | $0.50$2.50 over 100K | $0.05 | $0.25 | 1M | Current | 2026-10-09 |
Claude Fable 5.2 claude-fable-5-2 | $8.00 | $0.20−98% | $40.00 | $4.00 | $20.00 | 1M | Flagship | 2026-11-09 |
Claude Mythos 5.2 claude-mythos-5-2 | $8.00 | $0.20−98% | $40.00 | $4.00 | $20.00 | 1M | Restricted · Verified organisations only - Anthropic's Cyber and Life Sciences Verification Programs; Claude Enterprise customers also reach it through Claude Security. No self-serve access | 2026-11-09 |
Claude Fable 5.1 claude-fable-5-1 | $10.00 | $0.25−98% | $50.00 | $5.00 | $25.00 | 1M | Current | 2026-10-29 |
Claude Mythos 5.1 claude-mythos-5-1 | $10.00 | $0.25−98% | $50.00 | $5.00 | $25.00 | 1M | Restricted · Glasswing programme only - vetted cybersecurity and life-sciences organisations. No self-serve API alias published | 2026-10-29 |
Claude Fable 5 claude-fable-5 | $10.00 | $1.00−90% | $50.00 | $5.00 | $25.00 | 1M | Current | 2026-10-29 |
Claude Mythos 5 claude-mythos-5 | $10.00 | $1.00−90% | $50.00 | $5.00 | $25.00 | 1M | Restricted · Glasswing partners first; broader trusted access planned. No self-serve API alias published in launch post | 2026-10-29 |
Claude Opus 5 claude-opus-5 | $5.00 | $0.50−90% | $25.00 | $2.50 | $12.50 | 1M | Current | 2026-10-29 |
Claude Sonnet 5 claude-sonnet-5 | $2.00 | $0.20−90% | $10.00 | $1.00 | $5.00 | 1M | Current | 2026-10-29 |
| Claude Opus 4.8 | $5.00 | $0.50−90% | $25.00 | $2.50 | $12.50 | 1M | Current | 2026-10-29 |
| Claude Opus 4.7 | $5.00 | $0.50−90% | $25.00 | $2.50 | $12.50 | 1M | Current | 2026-10-29 |
| Claude Opus 4.6 | $5.00 | $0.50−90% | $25.00 | $2.50 | $12.50 | 1M | Current | 2026-10-29 |
| Claude Sonnet 4.6 | $3.00 | $0.30−90% | $15.00 | $1.50 | $7.50 | 1M | Current | 2026-10-29 |
| Claude Opus 4.5 | $5.00 | $0.50−90% | $25.00 | $2.50 | $12.50 | 200K | Current | 2026-10-09 |
| Claude Haiku 4.5 | $1.00 | $0.10−90% | $5.00 | $0.50 | $2.50 | 200K | Current | 2026-10-09 |
| Claude Sonnet 4.5 | $3.00 | $0.30−90% | $15.00 | $1.50 | $7.50 | 200K | Deprecated · retires Nov 30, 2026 · → Claude Sonnet 5.5 | 2026-10-29 |
US dollars per million tokens, list rates on Anthropic's own API. Cache read is the price of a cached input token. Batch is the asynchronous API, where the vendor offers one. The date in each row is when we last read it on the vendor's pricing page; it links to that page.
Claude API monthly cost estimates
What light use (10M input + 2M output tokens a month), medium use (100M input + 20M output tokens a month), heavy use (1B input + 200M output tokens a month) costs on each current model, at list price with no caching and no batch discount.
| Model | Input / output per 1M | Light · 10M + 2M | Medium · 100M + 20M | Heavy · 1B + 200M |
|---|---|---|---|---|
| Claude Haiku 5.5* | $0.10 / $0.50 | $2.00 | $20.00 | $200 |
| Claude Haiku 4.5 | $1.00 / $5.00 | $20.00 | $200 | $2,000 |
| Claude Sonnet 5.5, 5 | $2.00 / $10.00 | $40.00 | $400 | $4,000 |
| Claude Sonnet 4.6 | $3.00 / $15.00 | $60.00 | $600 | $6,000 |
| Claude Opus 5.5 | $4.00 / $20.00 | $80.00 | $800 | $8,000 |
| Claude Opus 5, 4.8, 4.7, 4.6, 4.5 | $5.00 / $25.00 | $100 | $1,000 | $10,000 |
| Claude Fable 5.2 | $8.00 / $40.00 | $160 | $1,600 | $16,000 |
| Claude Fable 5.1, 5 | $10.00 / $50.00 | $200 | $2,000 | $20,000 |
* Assumes every prompt stays under 100K tokens; above that the model bills its higher long-prompt rates. Prompt caching and the Batch API cut these figures; the next sections show by how much. For your own token counts, use the LLM API cost calculator.
Claude prompt caching and Batch API pricing
Prompt caching bills a repeated prompt prefix, such as a long system prompt or a document, at a lower rate after the first request. You pay once to write the prefix into the cache, then a fraction of the input price each time a later request reads it.
Anthropic prices a cache write by how long it lives. A 5-minute write costs 1.25 times the input rate, and a 1-hour write costs twice the input rate. A cache read costs a tenth of the input rate on most Claude models. The newest models read for less: $0.20 on Opus 5.5, $0.10 on Sonnet 5.5 and $0.25 on Fable 5.1.
| Model | Input | 5-min write | 1-hour write | Cache read | Batch in | Batch out |
|---|---|---|---|---|---|---|
| Claude Haiku 5.5 | $0.10 | $0.125 | $0.20 | $0.01 | $0.05 | $0.25 |
| Claude Sonnet 5.5 | $2.00 | $2.50 | $4.00 | $0.10 | $1.00 | $5.00 |
| Claude Opus 5.5 | $4.00 | $5.00 | $8.00 | $0.20 | $2.00 | $10.00 |
| Claude Fable 5.2 | $8.00 | $10.00 | $16.00 | $0.20 | $4.00 | $20.00 |
US dollars per million tokens. Haiku 5.5 rates are for prompts up to 100,000 tokens.
A 5-minute write pays for itself the first time a later request reads the prefix. A 1-hour write needs two reads to come out ahead, so it suits a prompt reused across a long working session.
The Batch API processes requests asynchronously at half the standard input and output rates. It suits evaluations, bulk extraction and any job that does not need an answer within seconds. The glossary explains prompt caching and the batch API in more detail.
Claude API surcharges: long prompts, US-only inference and fast mode
Claude Haiku 5.5 is the first Claude model priced by prompt length. Up to 100,000 input tokens it bills $0.10 input and $0.50 output per million. Once a prompt passes 100,000 tokens, the whole request bills $0.50 input and $2.50 output. The count includes cache reads and writes.
The Claude API runs requests on global infrastructure by default. Setting inference_geo to us keeps processing in the United States, and on Claude Opus 4.6 and later models it multiplies every token price by 1.1.
The Claude API has no EU-only option. For EU data residency, use a regional endpoint on Amazon Bedrock or Google Cloud. Anthropic's pricing page says regional and multi-region endpoints there cost 10% more than global ones for Claude 4.5 and later models.
Fast mode buys faster output on the Opus line at a premium rate. Opus 5.5 in fast mode costs $8 input and $40 output per million tokens. Opus 5 fast mode costs $10 input and $50 output, and Opus 4.8 fast mode costs $10 and $50.
Claude subscription or API: which is cheaper?
A Claude plan and the Claude API are separate products with separate bills. A plan pays for the Claude apps and Claude Code with a usage allowance. The API bills every token at the rates in the table above and has no monthly fee.
| Plan | Price | Usage | API credit included |
|---|---|---|---|
| Free | $0 | Smallest allowance | None |
| Pro | $20 a month, or $17 billed yearly | More than Free | None |
| Max | $100 a month | 5x Pro | $100 a month |
| Ultra | $200 a month | 10x Pro weekly usage, no 5-hour limit for verified accounts | $200 a month |
Plan prices from our plan sheet, checked 2026-11-05 on claude.com/pricing.
Ultra is the plan Anthropic used to call Max 20x; the price did not change when it was renamed. Each step up is compared in Claude Pro vs Max and Claude Max vs Ultra.
Pay per token when your use is light or uneven. If a month of your work would cost less than $20 at API rates, the API is the cheaper choice. A plan wins when you work in the Claude apps or Claude Code every day, because the allowance is not metered in dollars and heavy daily use can cost more than the plan price at API rates. The subscription vs API calculator runs the numbers for your own volume.
Claude Code works with both. Signed in with Pro, Max or Ultra, it draws on the plan allowance. Run on a Console API key, it bills per token at the API rates. For the plans side by side, see Claude Pro vs Max.
Claude API free credits, rate limits and tool fees
New Claude API accounts receive a small amount of free credit for testing. After that, every request is billed at the rates on this page. You create an API key in the Claude Console at platform.claude.com. The model ID under each name in the price table is the string you pass as model.
Rate limits grow with your usage tier. Anthropic names the tiers Start, Build and Scale, and a higher tier allows more requests and tokens per minute. Custom limits and volume discounts are negotiated with Anthropic's sales team.
Some server tools bill on top of tokens. Web search costs $10 per 1,000 searches. Code execution includes 1,550 free hours a month, then costs $0.05 per hour. Claude Managed Agents adds $0.08 per session-hour.
Claude Opus 4.7 and later models use a newer tokenizer. Anthropic says it can produce about 30% more tokens for the same text, so compare costs on token counts from the model you plan to use. The token counter gives a first estimate.
Claude API price history
Claude API price changes and releases
Releases, price changes and retirements on the Claude API, newest first. Each entry comes from an Anthropic announcement or from its pricing and deprecation pages.
Retired Claude models and their replacements
11 Claude models no longer answer on Anthropic's API. The price is the last one Anthropic listed; the replacement is the current end of the vendor's own succession chain.
| Model | Status | Last price in / out per 1M | Use instead |
|---|---|---|---|
| Claude Opus 4.1 | Retired Aug 5, 2026 | $15.00 / $75.00 | Claude Opus 4.8 |
| Claude Opus 4 | Retired Jun 15, 2026 | $15.00 / $75.00 | Claude Opus 4.8 |
| Claude Sonnet 4 | Retired Jun 15, 2026 | $3.00 / $15.00 | Claude Sonnet 4.6 |
| Claude Sonnet 3.7 | Retired Feb 19, 2026 | $3.00 / $15.00 | Claude Sonnet 4.6 |
| Claude Sonnet 3.5 (October 2024) | Retired Oct 28, 2025 | $3.00 / $15.00 | Claude Sonnet 4.6 |
| Claude Sonnet 3.5 (June 2024) | Retired Oct 28, 2025 | $3.00 / $15.00 | Claude Sonnet 4.6 |
| Claude Haiku 3.5 | Retired Feb 19, 2026 | $0.80 / $4.00 | Claude Haiku 4.5 |
| Claude Opus 3 | Retired Jan 5, 2026 | $15.00 / $75.00 | Claude Opus 4.7 |
| Claude Sonnet 3 | Retired Jul 21, 2025 | $3.00 / $15.00 | Claude Sonnet 4.6 |
| Claude Haiku 3 | Retired Apr 20, 2026 | $0.25 / $1.25 | Claude Haiku 4.5 |
| Claude Mythos Preview | Withdrawn | $25.00 / $125 | — |
platform.claude.com - direct API access, workbench and billing, at the list prices on this page. Global routing by default; US-only inference costs 1.1x on newer models.
CLOUD · AWSClaude on Amazon Bedrock, billed by AWS. Global endpoints, plus regional endpoints that keep data in a region at a 10% premium for Claude 4.5 and later. Anthropic also runs Claude Platform on AWS.
CLOUD · GCPClaude on Google Cloud (Vertex AI). Global, multi-region and regional endpoints; multi-region and regional carry a 10% premium for Claude 4.5 and later.
CONSUMER · WEBPro, Max and Ultra plans for the Claude apps and Claude Code. They are billed apart from the API, but Max includes $100 of API credit a month and Ultra includes $200.
AGENT · CLIAnthropic's coding agent. Signed in with a Claude plan it uses the plan's allowance; run on a Console API key it bills per token at the rates on this page.
OpenAI
GPT-6.1 Sol lists at $2.00/$10.00 per M and GPT-6 Astra at $10.00/$50.00. OpenAI also sells image-generation and realtime-audio models, which Anthropic does not.
CLOUD GIANTGemini 3.8 Flash carries a 1M-token window at $0.75/$3.75 per M. Google also sells image-generation and native-audio models, which Anthropic does not.
EU CHAMPIONMistral
EU-based, with open weights for much of its range. Mistral Large 4, in public preview, lists at $1.36/$4.18 per M; Large 3, with downloadable weights, at $0.50/$1.50.
COST DISRUPTORDeepSeek
DeepSeek V4 Pro lists at $1.32/$3.96 per M with a 1M-token window - a fraction of Opus 5.5's $4.00/$20.00. China-based.
FRONTIER RIVALSpaceXSI (Grok)
Grok 4.7 lists at $2.00/$6.00 per M below a 200K-token prompt and doubles above it, with a 500K window.
OPEN-WEIGHT CHALLENGERMoonshot (Kimi)
Kimi K3 lists at $3.00/$15.00 per M with a 1M-token window.
Claude API pricing FAQ
Short answers on what Claude API usage costs. Every rate comes from the price table at the top of this page.
Q · 01 How much does the Claude API cost?
+
Q · 02 Is the Claude API free?
+
Q · 03 Is the Claude API included in Claude Pro, Max or Ultra?
+
Q · 04 What is the cheapest Claude model?
+
Q · 05 How much does Claude Opus 5.5 cost?
+
Q · 06 How much does Claude Sonnet 5.5 cost?
+
Q · 07 How much does Claude Fable 5.2 cost?
+
Q · 08 How much do prompt caching and the Batch API save on Claude?
+
Q · 09 Is the Claude API cheaper than OpenAI or Gemini?
+
Q · 10 Why is my Claude API bill higher than the list price?
+
Anthropic was founded in 2021 by Dario and Daniela Amodei alongside several former OpenAI researchers. The pitch: build frontier-grade AI with safety and interpretability as first-class concerns, not afterthoughts.
The company is structured as a public-benefit corporation (PBC), which legally requires its directors to weigh a stated public benefit alongside shareholder return. Its Responsible Scaling Policy, which ties model capabilities to required safeguards, reached version 3.0 on February 24, 2026.
Anthropic's best-known technical contributions are Constitutional AI (training models against explicit written principles) and the Model Context Protocol, the open standard for connecting models to tools and data, which Anthropic donated to the Linux Foundation's Agentic AI Foundation in December 2025.
In May 2026 Anthropic raised $65 billion in a Series H at a $965 billion post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital; its previous round was a $30 billion Series G at $380 billion. Amazon and Google resell Claude through Bedrock and Google Cloud, and Microsoft through Foundry.
The lineup is narrower than OpenAI's or Google's: Anthropic sells no image-generation model and no embedding model of its own - its documentation points to Voyage AI for embeddings - and concentrates on text-and-image-input models for coding, agents and knowledge work.