Claude Sonnet 5
Direct API pricing, with every rate traced to its source and observation date.
Every cost, in context.
Current verified rates from the direct provider API. Each line has its own unit and condition.
| Price dimension | Condition | Rate / unit | Evidence |
|---|---|---|---|
| STANDARDInput tokens | New input · standard API | $2.00USD / 1M tokens | Source ↗ |
| Output tokens | Generated output · standard API | $10.00USD / 1M tokens | Source ↗ |
| PROMPT CACHECache read | Eligible reused input | $0.20USD / 1M tokens | Source ↗ |
| Cache write | 5-minute retention · eligible prefix | $2.50USD / 1M tokens | Source ↗ |
| Cache write · 1 hour | 1-hour retention · eligible prefix | $4.00USD / 1M tokens | Source ↗ |
| BATCHBatch input | Asynchronous batch request | $1.00USD / 1M tokens | Source ↗ |
| Batch output | Asynchronous batch response | $5.00USD / 1M tokens | Source ↗ |
What changed, and when.
Observation dates belong to this registry. Provider effective dates appear separately when announced.
Tracking started 2026-10-16 10:56 UTC. This registry has one price point; it cannot establish whether the provider changed prices before that date.
- Tracking started
Input $2.00 · Output $10.00 · Blended $4.00
Provider effective date not announcedOfficial price source ↗
Unchanged checks do not create new points. This timeline tracks verified input and output tariffs; other rate dimensions are shown above at their current values.
Make the numbers
your own.
Start with 6,000 input and 400 output tokens per task. Edit the assumptions below.
Standard API · your stated volume
Shape the workload01—05
Every dollar, accounted for.
Same workload. Different tariff conditions.
Prices from the verified provider source ↗. Volumes and cache reuse are your assumptions. Cache requires provider eligibility and retention; this estimate excludes tools, retries, taxes and success rate. Batch is a separate scenario.
When reuse pays off.
First write plus subsequent reads of the same eligible prefix. Compare with uncached input for 1 million prefix tokens.
- 1 reuse5-minute cache · first saving
- 2 reuses1-hour cache · first saving
Reuses must occur before each cache tier expires. Provider minimum prefix length and other eligibility rules still apply. Output and new input cancel in this comparison.
More in Claude Sonnet.
Names and versions identify a line. They do not prove equal quality, speed or capability.
Other versions will appear here when their prices pass verification and publication review.
Check the source.
Every current rate is attached to a dated source snapshot.
- Official pricing page
- Open provider source ↗
- Last verified
- 2026-10-23 08:49 UTC
- Provider effective date
- Not announced in verified source
- Source snapshot SHA-256
edf4e6f8aec260f1783b94913f77f152663f6c236e79dd620af332389d274769