Theoretical spend across pricing tiers, by model, in USD or EUR
Token volumes are derived from your real ccusage data (CLI + cowork). Tier rates are applied as a flat % discount on the per-model standard cost. EUR rate is editable — defaults to 0.92 (typical mid-market 2025–2026). Verify against your bank's rate for actual conversion.
Compares your actual API spend against subscription plans, projected to a full month and full year. Uses your usage rate of . The cheapest viable option is highlighted; amber rows likely exceed plan limits for typical Claude Code workloads.
| Plan / Billing Mode | Cost / month | Cost / year | vs API Standard (yr) | Fit for your workload |
|---|
Token volume: ~25M cache reads + ~3M cache writes + ~250K output per month at current rate. Subscription plans meter by message count and time-windowed limits (e.g. "messages per 5h"), not by tokens — so high-volume Claude Code workloads typically exceed Pro and even Max 5× before reaching the equivalent dollar value. API billing has no such limits.
Verified against Anthropic API docs and claudelog.com/claude-code-pricing. Anthropic does not publish % volume discounts for the API — Pro/Max are fixed-fee subscriptions, not API tiers.
| Tier | Discount vs Standard | Source / Eligibility |
|---|---|---|
| No Cache | + ~265% | Hypothetical: workload re-billed without prompt caching (cache reads at full input price). Shows what caching saves you. |
| Standard | 0% | Default Anthropic API pricing — what Claude Code is billed at. Includes automatic prompt caching (cache reads at 10% of input). |
| Priority Tier | 0% (same price) | Anthropic's committed-throughput tier — same per-token price as Standard; pays for guaranteed capacity, not a discount. |
| Batch API | −50% | Anthropic Batch API — 50% off input + output (and cache types). Async, 24h SLA. Not used by Claude Code (interactive). |
Pro and Max are fixed monthly fees with bundled usage limits — not per-token API tiers. Compare against your scenario's API spend to see if a subscription would have been cheaper. Claude Code on a subscription uses bundled allowance instead of API billing.
| Plan | Price / month | Models | Usage |
|---|
| Model | Input | Output | Cache Write (5m, 1.25×) | Cache Write (1h, 2×) | Cache Read (10%) |
|---|---|---|---|---|---|
| Opus 4.5 / 4.6 / 4.7 | $5.00 | $25.00 | $6.25 | $10.00 | $0.50 |
| Opus 4.6 — >200K context | $10.00 | $37.50 | $12.50 | $20.00 | $1.00 |
| Sonnet 4.6 | $3.00 | $15.00 | $3.75 | $6.00 | $0.30 |
| Haiku 4.5 | $1.00 | $5.00 | $1.25 | $2.00 | $0.10 |