Enter a representative monthly workload. The calculator applies the same input and output volume to each model so the result shows the direct price difference. It is a planning baseline, not a substitute for measuring real production usage.
Pricing note: Sonnet 5 uses its introductory $2 input / $10 output rate through August 31, 2026. Its announced standard rate is $3 / $15. All rates are per million tokens. Confirm Anthropic's current model pricing before approving a budget.
20.0M input + 5.0M output tokens
What this estimate includes
The calculation is straightforward: monthly requests multiplied by average input tokens and output tokens, then multiplied by each model's base per-million-token rates. It lets a buyer compare the same workload across models without hiding the input/output price difference.
What this estimate excludes
- Prompt cache writes, one-hour cache writes, and cache reads.
- Batch API discounts for asynchronous workloads.
- Long-context premiums, fast mode, web search, code execution, and other feature charges.
- Retries, refusals, fallback requests, agent loops, tool-result context, and evaluation traffic.
- Cloud-provider or enterprise commercial terms.
- Engineering, review, monitoring, incident handling, and operational support.
Why token price is not total workflow cost
A lower-cost model can become more expensive when it needs repeated attempts, generates longer output, creates more review work, or fails expensive edge cases. A higher-cost model can be economical when it completes difficult work in fewer steps and reduces human correction. Compare models on accepted outcomes per dollar, not tokens per dollar.
Build an evaluation set, measure accepted completion rate, capture actual token usage, and include review time. Datrick's production AI workflow guide covers the controls required around the model.
Frequently asked questions
How is Claude API cost calculated?
Base Claude API cost is calculated separately for input and output tokens: input tokens divided by one million multiplied by the model's input rate, plus output tokens divided by one million multiplied by the output rate. Production bills can also include prompt caching, batch discounts, tools, long-context premiums, and other features.
Which Claude model is cheapest?
Among the latest generally available Claude models, Haiku 4.5 has the lowest base token price at $1 per million input tokens and $5 per million output tokens as of July 2026.
Does the calculator include prompt caching or batch discounts?
No. It intentionally shows base input and output token cost so models can be compared consistently. Prompt caching, Batch API discounts, long-context pricing, fast mode, and platform-specific charges should be modeled separately for a production forecast.
Why can the real Claude API bill differ from this estimate?
Actual token counts vary by prompt, model tokenizer, retrieved context, tool results, retries, thinking behavior, and response length. Prices and temporary promotions also change. Measure real usage and confirm Anthropic's current pricing before approving a budget.
Need an accepted-outcome cost comparison? Datrick can benchmark your prompts and data across candidate models, then design routing, fallback, monitoring, and budget controls.
Request a model cost assessment