Claude Sonnet 4.5 API pricing at a glance
Anthropic lists Claude Sonnet 4.5 at $3 per million input tokens and $15 per million output tokens on its API. Prices below are in USD. This is API token billing, separate from Claude consumer subscription plans. (Anthropic pricing; Claude plans and pricing)
| Anthropic API usage | Price per million tokens (USD) |
|---|---|
| Input | $3.00 |
| Output | $15.00 |
| Prompt-cache write, 5-minute duration | $3.75 |
| Prompt-cache write, 1-hour duration | $6.00 |
| Cache read/hit | $0.30 |
| Batch API input and output | 50% discount on each standard rate |
Anthropic’s model reference labels the cache-write durations and lists the Batch API discount for input and output. (Sonnet 4.5 model reference)
What about requests that need more than 200K tokens?
Anthropic’s direct-API reference lists Sonnet 4.5 with a 200K-token context window and a 64K-token maximum output, alongside the $3/$15 standard rates. The sources cited here do not establish whether extended-context access is available for this model or, if it is, what access conditions or rates apply. If your workload requires more than the documented 200K context, confirm availability and billing with the provider before budgeting; don’t assume a higher rate or that the option is available. Hosting through another provider may have separate prices, limits, or lifecycle terms. (Sonnet 4.5 model reference; Anthropic pricing)
Estimate your Claude 4.5 API bill
For standard, uncached API usage:
Estimated token charges = (input tokens ÷ 1,000,000 × $3) + (output tokens ÷ 1,000,000 × $15)
For example, if a month’s requests use 1 million input tokens and 200,000 output tokens, the standard-rate estimate is:
- Input: 1 × $3 = $3.00
- Output: 0.2 × $15 = $3.00
- Estimated total: $6.00
This is a token-charge example, not a measured bill. It assumes the requests fit the model’s documented context limit and use standard, non-batch, uncached processing. It excludes cache writes and reads, batch processing, and any other separately billed items that may apply to your setup. (Anthropic pricing; Sonnet 4.5 model reference)
Production limits and model status
Anthropic documents text and image inputs, text output, a 200K context window, and up to 64K output tokens for Sonnet 4.5. Those listed limits help determine whether a planned workload fits; they are not separate prices. (Sonnet 4.5 model reference)
As of November 9, 2026, Anthropic marks Sonnet 4.5 deprecated and lists November 30, 2026 as its retirement date. Anthropic’s model reference recommends Claude Sonnet 5.5 as the replacement. If you already use Sonnet 4.5, plan for migration; for a new production service, compare the currently supported model’s pricing and capabilities before committing. (Sonnet 4.5 model reference)
Sources
Pricing — Claude Platform Docs