Worked example
100K input + 2K output
$0.11
Estimated token cost for one request. At 10,000 identical requests per month, the token subtotal would be $1,100.
Open this scenario →Anthropic pricing profile
Claude Haiku 4.5
Fast Claude model for high-volume and latency-sensitive steps.
active
economy
200,000 context
Verified 2026-08-02
Pricing modes
USD per one million tokens. Unsupported rate types are shown as unavailable rather than treated as zero.
| Mode | Input | Cached input | Cache write | Output | Long-context band |
|---|---|---|---|---|---|
| standard | $1 | $0.1 | $1.25 | $5 | — |
| batch | $0.5 | — | — | $2.5 | — |
Technical limits
Model constraints
Context window200,000 tokens
Maximum output64,000
Long-context thresholdNot applicable
Tokenizer profileanthropic_legacy
What can change the invoice
- Cached-token savings depend on provider eligibility and correct cache implementation.
- Taxes, cloud marketplaces, enterprise contracts, retries, latency and quality differences are not included in this example.
Verification record
Pricing was checked against first-party documentation on 2026-08-02. The active schedule is effective from 2025-01-01.
Related Anthropic models
Compare models from the same provider before evaluating quality and latency.
Claude Fable 5
Highest-tier Claude model for long-running, capability-first workflows.
View pricing →Claude Opus 5
Advanced Claude model for complex agentic coding and enterprise work.
View pricing →Claude Sonnet 5
Production-focused Claude model with temporary introductory pricing through August 2026.
View pricing →