Claude API Pricing 2026: Per-Token Rates, Real Cost Examples, and Free API Credits

Every Claude model's per-token rate, the multipliers that change your bill more than the list price, three worked examples with real numbers, and how startups get the API credits that make all of it free.

Haiku 4.5 $1 / $5 Input / output per MTok

Sonnet 5 $2 / $10 Most production workloads

Opus 5 $5 / $25 Complex multi-step work

Startups Free Credits via Claude for Startups

Claude API pricing per model (September 2026)

All prices in USD per million tokens, from Anthropic's pricing docs. Cache write columns are what you pay to store a prefix; cache read is what you pay when a later request reuses it.

Model	Input	Output	5-min cache write	1-hour cache write	Cache read
Claude Fable 5.1	$10	$50	$12.50	$20	$0.25
Claude Fable 5	$10	$50	$12.50	$20	$1
Claude Opus 5	$5	$25	$6.25	$10	$0.50
Claude Opus 4.8 / 4.7 / 4.6 / 4.5	$5	$25	$6.25	$10	$0.50
Claude Sonnet 5	$2	$10	$2.50	$4	$0.20
Claude Sonnet 4.6 / 4.5	$3	$15	$3.75	$6	$0.30
Claude Haiku 4.5	$1	$5	$1.25	$2	$0.10

Older models (Opus 4.1, Opus 4, Sonnet 4, Haiku 3.5) are retired on the first-party API and remain only on Bedrock or Google Cloud at their old rates. Sonnet 5's $2/$10 launched as introductory pricing; Anthropic has confirmed it is now the standard rate and the planned increase to $3/$15 will not happen.

Batch API pricing (50% off)

The Batch API processes requests asynchronously, typically within hours, at half price on both input and output. It stacks with prompt caching. Fast mode is not available in batch.

Model	Batch input / MTok	Batch output / MTok
Claude Fable 5.1 / Fable 5	$5	$25
Claude Opus 5 / 4.8 / 4.7 / 4.6 / 4.5	$2.50	$12.50
Claude Sonnet 5	$1	$5
Claude Sonnet 4.6 / 4.5	$1.50	$7.50
Claude Haiku 4.5	$0.50	$2.50

Batch is the single biggest lever for offline workloads: nightly classification, document processing, evaluation runs, embeddings prep. If a request does not need an answer in seconds, it belongs in a batch.

$100M for Startups

Fuel your runway with our $100M commitment. Claim up to 2 years free with Zendesk for Startups

Featured Opportunity

The multipliers that change your bill

List price is the starting point. These are the modifiers Anthropic applies, and most production bills are shaped more by them than by the model choice.

Prompt caching

A 5-minute cache write costs 1.25x the input price; a 1-hour write costs 2x. Every read of that cached prefix costs 0.1x input (0.025x on Fable 5.1). So caching pays for itself after one read on the 5-minute TTL and after two reads on the 1-hour TTL. For any app with a long system prompt or shared context, caching is the difference between a viable margin and not.

Long context

Claude 4.6 and later include the full 1M-token context window at standard pricing. A 900k-token request is billed at the same per-token rate as a 9k-token request. There is no long-context surcharge.

Tokenizer change

Claude 4.7 and later use a newer tokenizer that produces roughly 30% more tokens for the same text. Sonnet 5 at $2/$10 is still cheaper than Sonnet 4.6 at $3/$15 after that, but compare on your real prompts rather than on list price.

Fast mode

Fast mode on Opus 5 and Opus 4.8 gives significantly faster output at $10 input / $50 output, double the standard Opus rate. It stacks with caching and data residency multipliers and is first-party API only.

Data residency

US-only inference (inference_geo: "us") on 4.6+ models adds a 1.1x multiplier to every token category. Global routing, the default, uses standard pricing. On Bedrock and Google Cloud, regional endpoints carry a 10% premium over global ones.

Tool use and server-side tools

Tool definitions count as input tokens, and each model adds a tool-use system prompt of a few hundred tokens. Web search is $10 per 1,000 searches plus tokens. Web fetch is free beyond tokens. Code execution is free when used with web search or web fetch; otherwise each org gets 1,550 free container-hours a month, then $0.05 per hour.

Managed Agents

Claude Managed Agents bill tokens at the rates above plus $0.08 per session-hour of runtime while the session is in the running state. Batch discounts do not apply to agent sessions.

Claude API cost: three worked examples

Real arithmetic using the rates above. Token counts are typical for each workload; substitute your own.

Workload	Model	Assumptions	Monthly cost (list)	With caching / batch
Support ticket triage, 10,000 tickets/month	Haiku 4.5	~3,700 tokens per conversation (Anthropic's own example)	~$37	~$19 in batch
RAG chatbot, 100,000 queries/month	Sonnet 5	5,000 input tokens (4,000 shared context + 1,000 query), 500 output tokens per query	Input 500M × $2 = $1,000 Output 50M × $10 = $500 $1,500	Cache the 4,000-token prefix: 400M × $0.20 = $80, plus 100M × $2 = $200 uncached, plus $500 output = $780
Coding agent, one developer, 20 sessions/month	Opus 5	50,000 input and 15,000 output tokens per one-hour session (Anthropic's worked example)	$0.625 per session $12.50	With 40,000 of the input tokens as cache reads: $0.445 per session, $8.90. Add $0.08/hour runtime if on Managed Agents.

The RAG row is the one to study. Same model, same traffic, half the bill, from one cache_control field. Before you switch models to save money, switch on caching.

Rate limits, tiers and volume pricing

Rate limits are set by usage tier, not by plan. Anthropic describes three standard tiers: Start (entry-level limits), Build (increased limits for growing apps) and Scale (the highest standard limits for production). You move up by spending and by time on the platform, and you can request limits beyond Scale through sales.

Volume discounts exist but are negotiated case by case, and Anthropic notes that academic and research discounts may be available. Enterprise customers get custom rate limits, dedicated support and invoicing. If you are spending five figures a month, ask.

Billing is monthly in arrears on actual usage, by credit card or invoice, and every dollar is visible in the Claude Console. If you buy through AWS Marketplace or Azure Marketplace, usage is metered in Claude Consumption Units at $0.01 each, converted from the same per-token rates after any negotiated discount.

Claude for Startups members get Anthropic's highest rate limits from day one, which is often worth as much as the credits themselves when you are launching.

For founders

How startups get Claude API credits free

Anthropic's Claude for Startups program is the direct route to free API credits. It is self-serve: a Claude Console account on your company domain, a company email, your website and a brief description of what you are building. The form at claude.com/programs/startups takes about two minutes.

See eligibility & how to apply Apply

The fine print

You qualify if your company was founded within the last four years and has not received Anthropic startup credits before. Equity funding from institutional investors, or backing by an Anthropic partner VC or accelerator, unlocks the larger credit allocations and priority rate limits. Members also get early access to launches, community events in six cities, and access to Claude Code, the Claude API, Claude Managed Agents and Claude Cowork.

Two things to plan around. The credits apply to the first-party API via the Claude Console only, not Bedrock, Vertex AI or Foundry. And when they run out, standard pricing kicks in automatically with no migration. Build your unit economics on the rate table above from the first day, and use the credit period to get caching and batching right.

For inference you run through a cloud provider, AWS Activate, Google Cloud for Startups and Microsoft for Startups credits all cover Claude on their respective platforms.

Claude for Startups

Up to $100,000

$25,000 to $100,000+ in Claude API credits (partner tier)
Priority rate limits and early access to new models
Access to Claude Code, Claude API, Managed Agents and Cowork

Step by step: from zero to credited Claude API usage

Order matters. Set up the Console before you apply so the credits have somewhere to land.

1

Create a Claude Console organization on your company domain

Sign up at platform.claude.com with a work email. New accounts receive a small amount of free credit to test with. Our guide to getting an Anthropic API key walks through keys, workspaces and spend limits.
2

Apply to Claude for Startups

Submit the form at claude.com/programs/startups. Describe specifically what Claude does in your product. Check our Anthropic program page for eligibility and common rejection reasons first.
3

Turn on prompt caching before you scale

Add a single top-level cache_control field for automatic caching, or place explicit breakpoints. Cache reads at 10% of input price are the largest saving available.
4

Move offline work to the Batch API

Anything that can wait hours gets 50% off. Classification, enrichment, evaluations and backfills all belong here.
5

Set workspace spend limits

Create a workspace per environment in the Console and cap spend so a runaway loop cannot burn the credits overnight.
6

Model your post-credit cost now

Use the worked examples above with your own token counts. When the credits end, standard rates apply automatically. Route overflow through Bedrock or Vertex AI on cloud credits if you have them.

Questions founders ask

How much does the Claude API cost?

Per million tokens: Haiku 4.5 is $1 input / $5 output, Sonnet 5 is $2 / $10, Opus 5 and Opus 4.5 through 4.8 are $5 / $25, and Fable 5.1 is $10 / $50. Cache reads cost 10% of the input rate and the Batch API is 50% off.

How many words is a million tokens?

Roughly 750,000 English words, since one token is about four characters or 0.75 words. Claude 4.7 and later use a tokenizer that produces about 30% more tokens for the same text, so measure on your own prompts.

Is there a free tier for the Claude API?

New Claude Console accounts receive a small amount of free credit to test the API. Beyond that, free Claude API credits come from Claude for Startups for companies founded in the last four years, and from cloud startup programs when you run Claude on Bedrock, Vertex AI or Foundry.

How do I get Anthropic API credits for my startup?

Apply to Claude for Startups at claude.com/programs/startups with a Claude Console account, a company email, your website and a short description of what you are building. Companies founded within the last four years with no prior Anthropic credits are eligible; institutional or partner-VC backing unlocks larger allocations.

What is the cheapest Claude model?

Claude Haiku 4.5 at $1 input / $5 output per million tokens, or $0.50 / $2.50 in batch. For most production workloads Sonnet 5 at $2 / $10 is the better default; Haiku is for high-volume, latency-sensitive tasks.

Does prompt caching really save money?

Yes. A cache read costs 10% of the input price (2.5% on Fable 5.1). A 5-minute cache write costs 1.25x input, so caching is net positive after a single read. For a RAG app with a 4,000-token shared context, caching roughly halves the monthly bill in our worked example.

Can I use AWS or Google Cloud credits for the Claude API?

Yes, when you call Claude through Amazon Bedrock or Google Cloud Vertex AI. The usage bills to your cloud account at that provider's rates, where AWS Activate or Google Cloud startup credits apply. Claude for Startups credits do not cover cloud-billed usage.