As of November 9, 2026, Anthropic’s current Sonnet model is Claude Sonnet 5.5. For the first-party Claude API, its standard rates are $2 per million input tokens and $10 per million output tokens (USD). The API model ID is claude-sonnet-5-5. These are API metering rates—not the price of a Claude app subscription. (platform.claude.com)
Claude Sonnet 5.5 API prices
| Usage type | Price per million tokens (USD) |
|---|---|
| Standard input | $2.00 |
| Standard output | $10.00 |
| 5-minute prompt-cache write | $2.50 |
| 1-hour prompt-cache write | $4.00 |
| Prompt-cache read | $0.10 |
Anthropic’s pricing documentation labels these units MTok, or one million tokens, and lists prices in USD. Output tokens cost five times as much per token as standard input tokens. (platform.claude.com)
Estimate a monthly API bill
Use this formula for standard, uncached usage:
Estimated cost = (input tokens ÷ 1,000,000 × $2) + (output tokens ÷ 1,000,000 × $10)
Example: Suppose a production app handles 10,000 requests a month. Each request uses an average of 500 input tokens and 300 billed output tokens. The output count should include any thinking tokens billed as output.
- Input: 10,000 × 500 = 5,000,000 tokens → $10
- Output: 10,000 × 300 = 3,000,000 tokens → $30
- Estimated total: $40 per month, or $0.004 per request
This is an illustrative estimate, not a guaranteed invoice. Substitute the token counts from your own workload. Claude’s thinking tokens are billed as output tokens, even when the thinking text is not returned, so actual output charges can be higher than the visible response alone suggests. (platform.claude.com)
Cache and batch pricing for production workloads
Prompt caching may reduce input costs when requests reuse the same prompt prefix. For Sonnet 5.5, Anthropic lists cache reads at $0.10 per million tokens, compared with $2 per million for standard input; cache writes cost more than ordinary input. Caching is most relevant when stable instructions or context are reused, and eligibility depends on the request and cache setup. (platform.claude.com)
Batch processing offers a 50% discount on input and output token prices for asynchronous work. At the listed base rates, that works out to $1 per million input tokens and $5 per million output tokens. It can lower costs for jobs that do not need an immediate response, but it is not the right choice for an interactive request that must return in real time. Anthropic’s documentation says prompt caching is available with the Message Batches API and that the two pricing discounts can stack; actual savings depend on using caching and batch processing for eligible requests. (platform.claude.com)
If a request specifies US-only inference through the relevant API setting, Anthropic lists a 1.1× price multiplier for token pricing categories. The standard rates above are the baseline; account for that uplift if your workload requires US-only inference. (platform.claude.com)
If you meant an older Claude Sonnet model
The model version matters: Sonnet prices are not interchangeable. Anthropic lists Sonnet 4.6 at $3 per million input tokens and $15 per million output tokens. Anthropic marks Sonnet 4.6 as a legacy model and identifies Sonnet 5.5 as the current Sonnet model. If your production request still uses Sonnet 4.6, use that model’s rate—not the Sonnet 5.5 rate—in your budget. (platform.claude.com)
API rates are separate from Claude subscriptions
These per-token rates apply to the Claude API. A Claude paid plan and Claude Console API usage are separate products with separate billing; don’t use the API calculation above to estimate the cost of using Claude in the app. If you access Claude through a cloud provider or marketplace rather than Anthropic’s own API, the provider invoices you, so confirm the rates and conditions for that channel. (platform.claude.com)
For a practical production budget, confirm the model ID, estimate input and billed output tokens from representative requests, then add any applicable cache, batch, or inference-location pricing. Recheck Anthropic’s rate card before relying on a forecast, since prices and model availability can change.
Sources
Pricing — Claude Platform Docs
Claude Sonnet 5.5 — Claude Platform Docs
Claude Sonnet 4.6 — Claude Platform Docs
Thinking — Claude Platform Docs
Batch processing — Claude Platform Docs
Prompt caching — Claude Platform Docs
Why is Claude API usage billed separately from my paid Claude plan?