Claude Haiku 5.5
Released on October 7, 2026
Changes from previous version
Successor to Claude Haiku 4.5. Context grows from 200K to 1M. Prompts up to 100K tokens are priced at 10% of Haiku 4.5's $1/$5 list rates. Terminal-Bench 4.0 rises from 0.0% to 39.2%, OSWorld 2.1 offline subset from 15.7% to 72.4%, and GDPval-AA v2.1 from 735 to 1620 Elo. Adds adjustable effort.
Release Summary
Anthropic's fastest, cheapest Haiku, for high-volume summaries, compaction, classification, live support, and subagents beside Claude Opus 5.5 and Claude Sonnet 5.5. First Haiku with adjustable effort. For prompts up to 100K tokens, list price is $0.10/M input and $0.50/M output, with cache reads at $0.01/M and cache writes at $0.125/M. Above 100K, those rates are $0.50, $2.50, $0.05, and $0.625. About 90% of requests to Claude Haiku 4.5 fell in the lower tier, and Anthropic says Haiku 5.5 costs around 75% less to run. Announcement scores: GDPval-AA v2.1 1620 Elo, AA-Briefcase v1.1 1578 Elo, OSWorld 2.1 offline subset 72.4%, Humanity's Last Exam 45.9% with no tools and 57.4% with tools, Terminal-Bench 4.0 39.2%, FrontierCode 1.1 Main 46.4%, and Chartography 46.4% with no tools. A measured price per task is $0.21 at max effort, $0.12 at xhigh, $0.08 at high, $0.05 at medium, and $0.02 at low, using the up-to-100K rates. API ID claude-haiku-5-5. 1M-token context, 128K max output, knowledge cutoff June 2026. Available on the Claude Platform, Claude apps, Claude Code, Amazon Web Services, Google Cloud, and Microsoft Azure. The same announcement cuts Claude Sonnet 5.5 cache reads from $0.20/M to $0.10/M.
Timeline
Claude Haiku 5.5 released
Anthropic launches Haiku 5.5 as claude-haiku-5-5 on the Claude Platform, Claude apps, Claude Code, and major clouds, with tiered pricing at $0.10/$0.50 up to 100K tokens.