Direct answer
Claude Sonnet 4’s listed API rates were $3 per million input tokens and $15 per million output tokens, in USD. But the original direct-API model, claude-sonnet-4-20250514, was retired on June 15, 2026. A July 8, 2026 provider notice says it was no longer available on the Anthropic Claude API after that date; Anthropic’s lifecycle documentation says retired-model requests fail. These rates are therefore a historical pricing reference—not a current quote for a model available to deploy through Anthropic’s direct API. (Provider notice; Anthropic model deprecations; Anthropic pricing)
| Claude Sonnet 4 API rate | Listed price (USD per 1 million tokens) |
|---|---|
| Input | $3 |
| Output | $15 |
Anthropic-operated and partner-operated services can have different lifecycle schedules. Check the specific provider if you mean a hosted service rather than Anthropic’s direct API. (Anthropic model deprecations)
Estimate token costs
For a historical estimate using the listed rates, price input and output separately:
Estimated token cost = (input tokens ÷ 1,000,000 × $3) + (output tokens ÷ 1,000,000 × $15)
For a hypothetical workload of 765,000 input tokens and 765,000 output tokens:
- Input: 765,000 ÷ 1,000,000 × $3 = $2.295
- Output: 765,000 ÷ 1,000,000 × $15 = $11.475
- Total: $13.77
This is arithmetic using the listed rates, not a current quote or a measured bill. It assumes those token volumes and excludes charges outside the input/output rates. Since the direct-API model has been retired, use the rates and model currently offered by your provider when planning a new deployment.
What production teams should check
- Confirm the provider and model ID. The original Sonnet 4 direct-API model is retired. Anthropic’s lifecycle guidance says partner-operated platforms set their own retirement schedules, so verify availability with the specific service you use. (Provider notice; Anthropic model deprecations)
- Recalculate using the model you will actually deploy. Do not treat the historical Sonnet 4 rates as a current price, or assume another model will perform identically for your workload. Test your production prompts and estimate from observed input and output token counts using that model’s applicable pricing.
- Keep model versions separate. A price or feature rule for Sonnet 4.6 or another version is not proof of the same rule for Sonnet 4.
Bottom line
Claude Sonnet 4’s listed historical rate was $3/MTok input and $15/MTok output (MTok means one million tokens). The original model was retired from Anthropic’s direct API on June 15, 2026. For a production budget, confirm the model and pricing available through your intended provider, then estimate from your workload’s actual token counts.
Sources
Pricing - Claude Platform Docs