Last Updated: November 13, 2026

Claude API pricing runs from $0.10 to $40 per million tokens in November 2026. Haiku 5.5 is the cheapest model at $0.10 input and $0.50 output per million tokens, Sonnet 5.5 costs $2/$10, Opus 5.5 costs $4/$20, and Claude Fable 5.2, released November 9, tops the range at $8/$40, according to Anthropic's official pricing documentation. That is 20% less than Fable 5.1, which Anthropic still sells at $10/$50.
This guide covers every current and legacy Claude model, how prompt caching and the Batch API cut a Claude bill, what long context, web search and tools add, three worked cost examples, and how Claude API pricing compares with OpenAI and Google Gemini.
How Much Does the Claude API Cost Per Model?
The Claude API costs $0.10 to $8 per million input tokens and $0.50 to $40 per million output tokens, depending on the model. Anthropic lists four current models on claude.com/pricing: Haiku 5.5, Sonnet 5.5, Opus 5.5 and Fable 5.2.
Model | Input (per 1M tokens) | Output (per 1M tokens) | Cache read (per 1M tokens) | Best for |
|---|---|---|---|---|
Haiku 5.5 (prompts up to 100K tokens) | $0.10 | $0.50 | $0.01 | High-volume classification, extraction, routing and chat |
Sonnet 5.5 | $2.00 | $10.00 | $0.10 | Default for coding, writing and agents |
Opus 5.5 | $4.00 | $20.00 | $0.20 | Complex reasoning and high-stakes accuracy |
Fable 5.2 | $8.00 | $40.00 | $0.20 | Frontier workloads |
Claude API prices are billed per million tokens processed. Input tokens are what you send (your prompt, context and documents), and output tokens are what the model writes back. On every current Claude model, output costs 5x the input rate, so long responses move a Claude bill more than long prompts do.
Which Claude Model Should You Use for Your Budget?
Sonnet 5.5 is the sensible default for most Claude API projects, Haiku 5.5 is the pick for high-volume simple tasks, and Opus 5.5 or Fable 5.2 should be reserved for work where a wrong answer costs more than the extra tokens.
Haiku 5.5 is Anthropic's budget model at $0.10 input and $0.50 output per million tokens for prompts up to 100,000 tokens. The real Claude bargain in late 2026 is not a flagship model but Haiku 5.5, which costs one twentieth of Sonnet 5.5 per token. Haiku 5.5 also costs a tenth of the older Haiku 4.5 ($1/$5), so teams still running Haiku 4.5 should test a switch.
Sonnet 5.5 costs $2 input and $10 output per million tokens, the same list price as Sonnet 5. Sonnet 5.5 cache reads cost $0.10 per million tokens, half of Sonnet 5's $0.20, so any Sonnet workload with repeated context gets cheaper by moving to the newer model.
Opus 5.5 costs $4 input and $20 output per million tokens, double Sonnet 5.5's rate. Opus 5.5 is also cheaper than the Opus models before it: Opus 5, Opus 4.8, Opus 4.7, Opus 4.6 and Opus 4.5 are all listed at $5/$25. Anthropic also sells a fast mode for Opus 5.5 at $8 input and $40 output per million tokens, which is not available through the Batch API.
Fable 5.2 is the most expensive current Claude model at $8 input and $40 output per million tokens, double the price of Opus 5.5. Fable 5.2 also has the deepest cache discount in the lineup: cache reads cost $0.20 per million tokens, or 2.5% of its standard input rate. Fast mode for Fable 5.2 costs $16 input and $80 output per million tokens, per Anthropic's pricing documentation.
What changed with Claude Fable 5.2?
Anthropic released Claude Fable 5.2 on November 9, 2026, at $8/$40 per million tokens, 20% below the $10/$50 that Fable 5.1 still costs, according to Anthropic's announcement. Anthropic says Fable 5.2 writes output up to 30% faster than Fable 5.1, and the model is available on the Claude API, Amazon Web Services, Google Cloud and Microsoft Azure.
For a team running Fable 5.1 today, the switch is a straight price cut. A workload that sends 10 million input tokens and receives 2 million output tokens a month costs $200 on Fable 5.1 ($100 input plus $100 output) and $160 on Fable 5.2 ($80 plus $80), a saving of $40 a month before caching.
Claude Mythos 5.2, the same underlying model as Fable 5.2 with fewer cyber and biology safeguards, carries the same $8/$40 list price. Anthropic limits Mythos 5.2 to organizations in its Cyber and Life Sciences Verification Programs.
Which older Claude models are still available, and what do they cost?
Anthropic still lists several older Claude models on its API pricing page, and most of them cost the same as or more than their newer replacements. Legacy prices per million tokens (input/output) are:
Fable 5.1 and Mythos 5.1: $10/$50
Fable 5 and Mythos 5: $10/$50
Opus 5, Opus 4.8, Opus 4.7, Opus 4.6 and Opus 4.5: $5/$25
Opus 4.1 and Opus 4: $15/$75
Sonnet 5: $2/$10
Sonnet 4.6, Sonnet 4.5 and Sonnet 4: $3/$15
Haiku 4.5: $1/$5
Tool use adds hidden input tokens to every Claude API call that enables tools. Anthropic's docs say tools add a system prompt of 286 tokens on Opus 5.5, Sonnet 5.5 and Haiku 5.5 (with tool choice set to auto or none), billed at the normal input rate. The tool definitions themselves also count as input, including tools Claude reaches through MCP servers.
How Much Do Prompt Caching and the Batch API Save on Claude?
Prompt caching cuts the cost of repeated input on Claude by 90% to 97.5%, and the Batch API cuts both input and output prices by 50% for jobs that can wait. Both discounts are published in Anthropic's pricing documentation.
Prompt caching stores a repeated block of input, such as a system prompt, reference document or codebase, so later Claude calls read it at a fraction of the normal price. Writing to the 5-minute cache costs 1.25x the base input price and writing to the 1-hour cache costs 2x. Cache reads cost 0.1x the base input price on most models, 0.05x on Opus 5.5 and Sonnet 5.5, and 0.025x on Fable 5.2 and Fable 5.1. On a model with 0.1x cache reads, the 5-minute cache pays for itself on the first reuse and the 1-hour cache on the second.
The Batch API gives a flat 50% discount on Claude input and output tokens for work that does not need an instant answer, such as bulk classification, overnight processing or large content jobs. Batch prices per million tokens are $1/$5 for Sonnet 5.5, $2/$10 for Opus 5.5, $4/$20 for Fable 5.2 and $0.05/$0.25 for Haiku 5.5 on prompts up to 100,000 tokens.
Caching and batching often change a Claude API bill more than model choice does. A 100,000-token document read from cache on Opus 5.5 costs $0.02 per call, against $0.40 at the standard input rate and $0.20 through the Batch API without caching.
Does Claude Charge Extra for Long Context, Web Search or Tools?
Claude does not charge a long-context premium on most current models, but Haiku 5.5, web search and some code execution carry extra costs. Anthropic says Claude 4.6 and later models, except Haiku 5.5, include the full 1M token context window at standard pricing, so a 900,000-token request is billed at the same per-token rate as a 9,000-token request.
Haiku 5.5 is the exception to flat long-context pricing on Claude. Once a Haiku 5.5 prompt passes 100,000 tokens, input rises from $0.10 to $0.50 per million tokens and output rises from $0.50 to $2.50, and cache reads and writes count toward that prompt length. A 150,000-token prompt with a 1,000-token answer costs about $0.08 on Haiku 5.5 at those rates, versus about $0.31 on Sonnet 5.5, so Haiku 5.5 stays cheaper even above the threshold.
Claude's server-side tools carry their own fees on top of token costs:
Web search: $10 per 1,000 searches, plus the standard token cost of processing results. Searches that return errors are not billed.
Web fetch: no extra fee; you pay only for the tokens in the fetched content.
Code execution: free when used with the current web search or web fetch tools. Otherwise each organization gets 1,550 free hours a month, then pays $0.05 per hour per container, with a 5-minute minimum.
US-only inference: keeping inference in the US adds a 1.1x multiplier on Claude 4.6 and later models.

What Does the Claude API Cost in Practice? Three Worked Examples
The worked examples below use Anthropic's published November 2026 rates. Each one shows how model choice, caching and batching change the final Claude API bill.
Customer support chatbot. A conversation with a 500-token question, 300 tokens of context and a 200-token reply uses 800 input and 200 output tokens. On Sonnet 5.5 that is $0.0016 + $0.002 = $0.0036 per conversation, or $36 for 10,000 conversations a month. On Haiku 5.5 the same 10,000 conversations cost $1.80.
Coding agent with a cached codebase. An Opus 5.5 agent makes 500 calls, each sending the same 100,000-token codebase plus 2,000 new input tokens and getting 1,000 output tokens back. Without caching the run costs about $214 ($204 input, $10 output). With the codebase cached, assuming the cache stays warm, the run costs about $24.50: $0.50 for one cache write, about $10 in cache reads, $4 for new input and $10 for output. That is roughly an 89% saving.
Bulk document summaries. Summarizing 50,000 documents at 3,000 input and 300 output tokens each means 150 million input and 15 million output tokens. Opus 5.5 at standard rates costs $900. Sonnet 5.5 through the Batch API costs $225. Haiku 5.5 through the Batch API costs $11.25.
Across all three Claude API examples, the model you pick and your use of caching or batching move the bill far more than small differences in headline rates. The same summary job ranges from $11.25 to $900 depending on those two decisions.
How Much Do 1 Million, 10 Million and 50 Million Claude Tokens Cost?
One million input tokens cost $0.10 on Haiku 5.5, $2 on Sonnet 5.5, $4 on Opus 5.5 and $8 on Fable 5.2, and output tokens cost five times as much on each model. The table below scales Anthropic's November 2026 list prices to common volumes, before caching or batch discounts.
Model | 1M input | 10M input | 50M input | 1M output | 10M output | 50M output |
|---|---|---|---|---|---|---|
Haiku 5.5 | $0.10 | $1 | $5 | $0.50 | $5 | $25 |
Sonnet 5.5 | $2 | $20 | $100 | $10 | $100 | $500 |
Opus 5.5 | $4 | $40 | $200 | $20 | $200 | $1,000 |
Fable 5.2 | $8 | $80 | $400 | $40 | $400 | $2,000 |
Haiku 5.5 figures assume each request's prompt stays under 100,000 tokens; above that, Haiku 5.5 input costs $0.50 and output $2.50 per million. The Batch API halves every figure in the table.
Token counts are not word counts. Anthropic says Claude 4.7 and later models use a newer tokenizer that produces roughly 30% more tokens for the same text than Sonnet 4.6 and earlier models, with the exact increase depending on the content (Anthropic). Because the newer tokenizer turns the same text into roughly 30% more tokens, a switch from Sonnet 4.6 to Sonnet 5.5 cuts the real bill by about 13%, not the 33% the rate card suggests. Measure tokens on your own prompts before budgeting a migration.
How Do You Estimate Your Monthly Claude API Cost?
To estimate a monthly Claude API bill, multiply your average input and output tokens per call by the model's rates, then multiply by your expected monthly call volume. Run the math on real prompts rather than guessing from the rate card.
[(Average input tokens per call × input rate) + (average output tokens per call × output rate)] × monthly call volume = estimated monthly cost
Here is a worked Claude API estimate. A feature that summarizes support tickets on Sonnet 5.5 averages 800 input tokens (the ticket plus instructions) and 150 output tokens per call, and runs 20,000 times a month. That is (800 × $0.000002) + (150 × $0.00001) = $0.0016 + $0.0015 = $0.0031 per call, or about $62 a month. Prompt caching helps little here because each ticket is unique.
An internal tool that answers questions against the same 50-page policy document all day is the opposite case for Claude API costs. The policy document is a large, unchanging block of context, which is exactly what prompt caching is built for: pay the cache-write premium once, then pay a fraction of the input rate on each call that reuses it. Knowing which type of workload you have before you build is what separates an accurate Claude cost estimate from a surprise bill. Our guide to how LLMs work explains the token mechanics behind this pricing model.
What Are the Most Common Claude API Pricing Mistakes?
Most unexpectedly high Claude API bills come from four habits: using Opus for everything, letting outputs run long, skipping the Batch API and resending the same context without caching.
Defaulting to Opus for everything. Opus 5.5 costs double Sonnet 5.5's rate, and for drafting, classification and straightforward Q&A, Sonnet 5.5 is often good enough that the extra cost buys little. Reserve Opus 5.5 for the calls where accuracy justifies it.
Ignoring output length. Because output tokens cost roughly 5x input tokens across the board, an unconstrained prompt that lets the model ramble costs meaningfully more than one with a clear length instruction or max-token cap. This is one of the cheapest optimizations available and one of the most commonly skipped.
Not using the Batch API for anything that can wait. Overnight processing, bulk analysis and other non-interactive Claude jobs that run synchronously instead of through the Batch API leave a flat 50% discount unused.
Resending the same large context on every call. Without caching, a workflow that repeatedly sends the same system prompt, reference document or codebase pays full input price on that content every time. Caching is usually the biggest cost lever once a Claude workload is live. For a broader look at where AI projects overspend, see our AI pricing guide for 2026.
Is the Claude API Cheaper Than OpenAI and Gemini?
Claude API list prices now undercut OpenAI's at the top of the range, match OpenAI's in the middle and at the bottom, and match Google's Gemini 4 Argon until Google's price increase on December 1, 2026. The other differences show up in caching rates and in tiers OpenAI and Google sell that Claude does not.
Model | Provider | Input (per 1M tokens) | Output (per 1M tokens) | Note |
|---|---|---|---|---|
Fable 5.2 | Anthropic | $8.00 | $40.00 | Cache reads $0.20 |
GPT-6 Astra | OpenAI | $10.00 | $50.00 | Cached input $1.00 |
Opus 5.5 | Anthropic | $4.00 | $20.00 | Cache reads $0.20 |
Sonnet 5.5 | Anthropic | $2.00 | $10.00 | Cache reads $0.10 |
GPT-6.1 Sol | OpenAI | $2.00 | $10.00 | Cached input $0.10 |
Gemini 4 Argon | $2.00 | $10.00 | Rises to $4/$20 on Dec 1, 2026 | |
GPT-6.1 Terra | OpenAI | $1.00 | $5.00 | Cached input $0.05 |
Gemini 3.8 Flash | $0.75 | $3.75 | Rises to $1.50/$7.50 on Jan 1, 2027 | |
Haiku 5.5 | Anthropic | $0.10 | $0.50 | Prompts up to 100K tokens |
GPT-6 Luna | OpenAI | $0.10 | $0.50 | Cached input $0.01 |
The prices above are standard per-million-token rates from the Anthropic, OpenAI and Google Gemini API pricing pages. All three were checked on November 12, 2026, with Google's Gemini rates and scheduled increases confirmed on Google Cloud's Vertex AI pricing page. For each provider's full rate card, see our OpenAI API pricing and Gemini API pricing guides. If you buy OpenAI models through Microsoft instead, see our Azure OpenAI pricing guide. If you run Claude on AWS, see our Amazon Bedrock pricing guide.
At the frontier, Fable 5.2 costs $8/$40 per million tokens against $10/$50 for GPT-6 Astra, and Fable 5.2 cache reads cost $0.20 per million tokens against $1.00 for Astra's cached input. Since November 9, Claude's top model has been 20% cheaper than OpenAI's on list price and far cheaper on reused context.
In the middle tier, Sonnet 5.5, GPT-6.1 Sol and Gemini 4 Argon all cost $2/$10 today, so the support chatbot example above costs $36 a month on any of them. After Google's December 1 increase, the same 10,000 conversations on Gemini 4 Argon cost $72, which puts Argon at Opus 5.5's list price.
At the budget end, Haiku 5.5 and GPT-6 Luna tie at $0.10/$0.50 with matching $0.01 cache reads. Gemini 3.8 Flash costs 7.5x more per token than Haiku 5.5 at $0.75/$3.75. OpenAI's GPT-6.1 Terra at $1/$5 fills a slot between Haiku 5.5 and Sonnet 5.5 that Claude does not cover.
Teams weighing other low-cost providers can also compare our breakdowns of DeepSeek pricing and Grok AI pricing. Our monthly AI Pricing Index tracks API rates across the major providers.
Is the Claude API Separate From a Claude Pro or Max Subscription?
Yes. The Claude API and Claude subscriptions are billed separately and solve different problems. Claude Pro costs $20 a month ($17 a month billed annually), Max costs $100 a month and Ultra costs $200 a month, per Anthropic's plans page. Those plans cover using Claude yourself in the chat apps or Claude Code, with usage limits. The API is for building Claude into your own product, billed per token with no subscription. How those subscription caps work is covered in our guide to Claude usage limits.
For an individual using Claude for their own work, a subscription is usually cheaper and simpler than the API. For an application that other people or systems call programmatically, the API is the right tool. Some teams pay for both: subscriptions for the people doing the work and API credits for the products they build. Claude Enterprise blends the two, at $20 per seat per month billed annually with a 20-seat minimum, plus usage at API rates.

Free download: The AI Stats Sheet 2026. 36 key AI numbers for business on four printable pages, with the source under every figure. Join 2,000+ business leaders who read AI Business Weekly and get it in your inbox right away.
Frequently Asked Questions
How much does the Claude API cost per month?
The Claude API has no fixed monthly cost. Anthropic bills per million tokens, so a Claude bill depends on volume and model: 10,000 short support conversations cost about $36 a month on Sonnet 5.5 and under $2 on Haiku 5.5, while large production systems can run into thousands of dollars.
Is Claude API pricing cheaper than ChatGPT's API?
Claude API pricing is level with or below OpenAI's at matching tiers in November 2026. Sonnet 5.5 and GPT-6.1 Sol both cost $2/$10 per million tokens, Haiku 5.5 and GPT-6 Luna both cost $0.10/$0.50, and Fable 5.2 costs $8/$40 against $10/$50 for GPT-6 Astra, with cheaper cache reads. Our Claude vs. ChatGPT comparison covers how the two differ beyond price.
What is the cheapest Claude API model?
Haiku 5.5 is the cheapest Claude API model at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. Through the Batch API, Haiku 5.5 drops to $0.05 input and $0.25 output.
Does prompt caching actually save significant money?
Yes, prompt caching saves significant money when an application resends the same context. Claude cache reads cost 10% of the standard input rate on most models, 5% on Opus 5.5 and Sonnet 5.5, and 2.5% on Fable 5.2 and Fable 5.1. A one-off query with no repeated context gets no benefit from caching.
Do I need a Claude Pro subscription to use the API?
No. Claude API access is billed per token through Anthropic's developer platform with no subscription required, while Pro, Max and Ultra are consumer plans for using Claude directly in chat or Claude Code. For more on Claude's products and growth, see our Claude AI statistics.
How much do 10 million Claude tokens cost?
Ten million input tokens cost $1 on Haiku 5.5, $20 on Sonnet 5.5, $40 on Opus 5.5 and $80 on Fable 5.2 at November 2026 list prices. Ten million output tokens cost $5, $100, $200 and $400 on the same models, and the Batch API halves those figures.
The Bottom Line on Claude API Pricing
Claude API pricing in November 2026 spans four current models, from Haiku 5.5 at $0.10/$0.50 to Fable 5.2 at $8/$40 per million tokens, with output always priced at 5x input. Prompt caching and the Batch API decide the final Claude bill as much as model choice does. Start with Sonnet 5.5, test Haiku 5.5 on simple high-volume tasks, measure real token usage, and add caching before chasing the cheapest sticker price.
⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.
Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.
