Claude Haiku 5.5

About the model

• Context: 1M tokens. Publisher limit, with up to 128K output tokens.
• Reasoning: Adaptive thinking. Medium effort is the publisher default.
• Best fit: Focused tasks. Try classification, extraction and summaries; larger models remain better suited to complex agentic coding.
• Migration: Review request options. Sampling settings, manual thinking budgets and assistant prefills have changed.

Anthropic publisher pricing

At up to 100,000 input tokens, input/output cost $0.10/$0.50 per million tokens. Above that threshold, the full request uses $0.50/$2.50 rates. Cache reads cost $0.01/$0.05 across the same tiers. These are publisher prices, before MiniRouter markup.

See current MiniRouter rates and how billing works.

Publisher-reported benchmarks

Results from the publisher’s announcement, grouped by unit.

Remove sampling parameters, manual thinking budgets and assistant prefills when migrating. Thinking consumes output tokens, so small output limits can finish before an answer appears.

Anthropic reports roughly 30% more input tokens than Haiku 4.5 for the same text. Recheck prompt budgets and compare completed-task cost.

Availability follows MiniRouter's public catalog. Publisher availability alone does not establish a working MiniRouter route.

Keep exploring

Compare published results in the model rankings and coding benchmarks, or read more model releases.

For MiniRouter product updates, visit the changelog.

Sources

1. Anthropic Haiku 5.5 announcement

   Published Oct 7, 2026

   Read at Anthropic ↗

2. Haiku 5.5 model overview

   Published Oct 7, 2026

   Read at Anthropic ↗

3. Haiku 5.5 migration guide

   Published Oct 7, 2026

   Read at Anthropic ↗

Editorial reference revised 2026-10-07. Catalog status comes from the generated routing snapshot.