Skip to main content

Claude cost comparison

Claude Sonnet vs Opus Cost: API Pricing Comparison

Claude Sonnet and Opus have different rate profiles and intended workloads. Compare them using the same token mix, then evaluate whether task quality changes retries or the number of model calls needed.

Formula-driven examplesSource-linked pricing snapshots

Output volume matters

Agentic coding and reasoning requests can produce substantial output and tool-call context. Because output has its own rate, a one-million-input-token comparison alone can understate the real difference.

Prompt caching may lower eligible repeated input, but cache writes and cache hits are different billing events. Model them separately when your application has stable prompt prefixes.

monthly difference = (Opus request cost − Sonnet request cost) × monthly requests

Measure cost per successful task

A more capable model can be economical when it reduces retries, human review, or multi-step calls. Run an evaluation set and compare successful-task cost rather than assuming either tier wins.

Worked example

Claude Sonnet 4.6: 1M input tokens + 1M output tokens

Using the versioned rates below, this example workload is estimated at $18.00. This isolates provider usage only and does not include taxes, regional premiums, retries, storage, network traffic, or unrelated infrastructure.

Current pricing references

These versioned records support the examples above. Check the date and provider source before using them in a production forecast.

Provider / modelInput or unitOutputStatusSource

Anthropic

Claude Sonnet 4.6

$3.00 per 1M tokens$15.00 / 1MVerifiedAnthropic Claude API pricing

Checked Jun 21, 2026

Anthropic

Claude Opus 4.8

$5.00 per 1M tokens$25.00 / 1MVerifiedAnthropic Claude API pricing

Checked Jun 21, 2026

Frequently asked questions

Which is cheaper, Claude Sonnet or Opus?

Sonnet has lower standard token rates in this snapshot. The better economic choice depends on output volume and success rate for your workload.

Does Claude batch pricing change the comparison?

It can for eligible asynchronous workloads. Compare both models under the same processing mode and confirm current Anthropic terms.

Related calculators and guides