Skip to main content
LogoHaiku5-5.com
  • Workspace
  • API calculator
  • Credit packs

Haiku 5.5 API pricing

These are manufacturer API estimates, not the price of using this website. Enter each model's input and billable output to compare a dated price snapshot.

Manufacturer API prices

Official cost calculator

Estimate direct API spending at published manufacturer rates. This is not a ZenMux bill or a charge to your site credits.

$0.000600

Anthropic source · 2026-10-09

$0.012000

Anthropic source · 2026-10-09

Baseline minus Haiku (USD)

$0.011400

95% difference. Negative means Haiku costs more.

Equal token counts assumed for both models. Billable output includes reasoning tokens; do not add them twice. Standard real-time pricing, excluding cache discounts, batch discounts and taxes. Long-input tiers apply to the whole call. Rates are dated snapshots, not live quotes.

Manufacturer reference

Official standard API specifications and dated prices. The verified ZenMux connection used by this site can have different limits and billing. No benchmark winner is implied.

ModelContext / max outputDefault effortInput / output (USD per 1M)
Claude Haiku 5.5Checked 2026-10-091,000,000 / 128,000medium
$0.10 / $0.50Standard tier
$0.50 / $2.50From 100,001 input tokens
Read the price as a rate, not a per-answer feeA long prompt can change both ratesReasoning belongs in billable outputCaching, batch and other exclusionsOfficial API cost, ZenMux cost and site creditsManufacturer sources

Read the price as a rate, not a per-answer fee

The table separates input from output and uses USD per million tokens. Multiply the input count by its rate, add billable output at its rate, then divide by one million. The calculator repeats that per-call estimate for the number of calls you enter.

The initial values are assumptions. They are not measured Haiku performance or an estimate of how many messages a credit pack buys. After a comparison, the usage panel can carry actual token counts into the calculator; fields you already edited remain unchanged.

A long prompt can change both rates

Haiku has a long-input threshold. The published price table shows exactly where the higher tier starts. A request above that boundary uses the higher input and output rates for the whole request, not just for the tokens beyond the boundary.

Count all input when checking the boundary, including cached portions where the manufacturer specifies them. Earlier calls do not change price retroactively because a later turn becomes longer. Each call has its own input length and applicable tier.

Reasoning belongs in billable output

Use the provider's billable output count rather than counting only the words visible in the answer. Reasoning can consume output tokens even when it is not shown in the conversation. If reasoning is already included in the reported output total, adding it again would overstate the estimate.

A missing usage value is unknown, not zero. The same rule applies to a supplier bill that has not arrived. Completed content can be available while its API cost still needs reconciliation.

Caching, batch and other exclusions

This calculator uses standard, real-time, uncached text rates. It does not model cache-write or cache-read discounts, the manufacturer Batch API, premium speed modes, regional processing charges, tax or paid server tools. Check the linked official pricing documentation for those cases.

A CSV batch on this website is a sequence of ordinary model calls. It does not imply that the manufacturer's discounted Batch API is being used. That distinction matters when estimating a large import.

Official API cost, ZenMux cost and site credits

Three different numbers can describe the same task. The calculator estimates a direct manufacturer API price. A saved report can show the actual ZenMux cost once reconciled. The credit ledger records the site charge using the rate version pinned at submission.

Credit packs are sold on the separate Credit packs page. They are one-time purchases with permanent batch access, not an Anthropic subscription. Refund questions go to [email protected] for individual review.

Manufacturer sources

  • Claude Haiku 5.5 model reference · Checked 2026-10-09
  • Anthropic API pricing · Checked 2026-10-09

These sources describe the manufacturers' products. This page does not publish a site benchmark or reproduce a third-party ranking. Your saved outputs are private observations of your own inputs and settings.

Continue your evaluation

Haiku 5.5Haiku 5.5 vs Sonnet 5.5Haiku 5.5 vs GPT-6 LunaHaiku 5.5 vs Opus 5.5Haiku 5.5 vs GPT-6.1 SolSite credit packsCompare and Chat workspace
LogoHaiku5-5.com

Independent model comparisons, grounded in your own tasks. Not affiliated with Anthropic or OpenAI.

[email protected]
Product
  • Compare & Chat
  • Credit packs
  • Input samples
  • Questions
Haiku 5.5
  • Model overview
  • Official API pricing
Comparisons
  • vs Sonnet 5.5
  • vs GPT-6 Luna
  • vs Opus 5.5
  • vs GPT-6.1 Sol
Policies
  • Privacy
  • Terms
  • Refunds
  • Cookies
© 2026 Haiku5-5.com. All Rights Reserved.