FREE COST TOOL • CLAUDE API PRICING
Claude API Cost
Calculator
Estimate first-party Anthropic Claude API spend by model, request volume, uncached tokens, 5-minute or 1-hour prompt-cache writes, cache reads, output, Batch or Fast processing, and global or US-only inference. This is API pricing—not a Claude Pro or Max subscription calculator.
EXAMPLE WORKLOAD · EDIT ANY FIELD
Visitor's hypothetical API workload—not your bill. This page is not connected to Anthropic, a card, or checkout and cannot charge anyone.
Checked July 24, 2026: Claude Sonnet 5 standard uses $2 uncached input, $2.5 5-minute cache writes, $4 1-hour cache writes, $0.2 cache reads, and $10 output per 1M tokens.
1,000 requests/month × $0.01035 · 2,800,000 total monthly tokens
$3 uncached input + $7 output
$0.25 writes + $0.1 reads
Versus billing all entered input and cache tokens as uncached input.
Same workload using published post-introductory Sonnet 5 rates.
US-only inference applies its multiplier to all token categories.
Compare the same workload across Claude models
| Model | Mode | Monthly estimate | Per request |
|---|---|---|---|
| Claude Haiku 4.5 | Standard | $5.18 | $0.00518 |
| Claude Sonnet 5 | Standard | $10.35 | $0.01035 |
| Claude Sonnet 4.6 | Standard | $15.53 | $0.01553 |
| Claude Opus 4.8 | Standard | $25.88 | $0.02588 |
Price is one constraint, not a quality ranking. Validate task accuracy, latency, context behavior, rate limits, and tool support on your own workload before routing production traffic.
WHAT OFTEN CHANGES THE INVOICE
Model the billing path, not just input × output.
- Prompt caching: separate 5-minute writes, 1-hour writes, and reads; they use different multipliers.
- Batch: use it only for asynchronous work that can accept Batch behavior and limits.
- US-only inference: the current 1.1× residency multiplier stacks with token and cache charges.
- Tools: web search, code execution, server tools, retries, and agent loops can add charges not modeled here.
- Tokenizer changes: measure actual usage; newer models can tokenize the same text differently.
- Long context: no separate over-200K premium is added for the current models shown here.
Frequently asked questions
Is Claude API included with Claude Pro or Max?
No. Claude.ai subscriptions and first-party Claude API usage are separate products and billing routes. This calculator estimates API token charges, not Pro, Max, Team, or Enterprise subscriptions.
How does Claude prompt caching affect cost?
A 5-minute cache write costs 1.25 times the base input rate, a 1-hour write costs 2 times, and a cache read costs 0.1 times. Reuse must be high enough for cheaper reads to offset the initial write.
How much cheaper is Claude Batch API?
Anthropic publishes a 50% Batch discount on input and output token processing. Prompt-cache and US-only inference multipliers can stack with Batch, so this calculator applies the selected conditions together.
When does Claude Sonnet 5 pricing change?
Anthropic lists introductory Sonnet 5 pricing through August 31, 2026, followed by higher rates from September 1, 2026. The calculator shows both estimates for the same workload.
Does Claude charge more above 200K context?
The current first-party pricing page says the supported current Opus and Sonnet models include the full 1M-token context window at standard pricing. This calculator does not add a separate long-context premium.
Does this include Amazon Bedrock or Google Vertex AI pricing?
No. This is an independent estimate for Anthropic's first-party Claude API. Partner-cloud rates, enterprise agreements, tools, web search, code execution, taxes, and credits can differ.
Independent estimate; not affiliated with Anthropic. Rates are public first-party Claude API list prices checked on July 24, 2026. This tool excludes Claude.ai subscriptions, partner clouds, Priority Tier, tool charges, web search, code execution, retries, taxes, negotiated discounts, prepaid-credit rules, and any charge outside the fields shown. Sonnet 5 introductory pricing changes on September 1, 2026. Verify the official pricing page and your invoice before purchase or launch.