Claude Haiku 5.5 is now available in Kiro

Claude Haiku 5.5 is available today across Kiro IDE, CLI, Web, and Crew. It is the fastest model in the Claude 5.5 family, aimed at work that runs in large volumes and needs a quick response, and it is a clear step up from Claude Haiku 4.5. With Haiku 5.5, all three Claude 5.5 models are now in Kiro. Opus 5.5 is there for your hardest problems and Sonnet 5.5 for everyday coding, while Haiku 5.5 steps in wherever volume or speed is what matters most.

Anthropic’s data shows Haiku 5.5 is the fastest model they have shipped so far, with significant quality gains over Haiku 4.5: 39.2% on Terminal-Bench 4.0 for agentic coding, up from 0.0%, and 1620 on GDPval-AA v2.1 for real-world knowledge work, up from 735.

Reach for Haiku 5.5 on high-volume, speed-sensitive work where cost matters: summarizing and compacting long context, classifying and routing requests, bulk text processing, and small, specific edits across many files. It also makes a strong review or computer-use subagent you can run in parallel.

It runs with Adaptive Thinking across five effort levels from low to max, defaulting to medium, and carries a 1M-token context window.

Haiku 5.5 ships with Anthropic’s safeguards, and its model controls are tighter than Haiku 4.5’s. They clear a broader set of defensive security tasks than the Sonnet 5.5 safeguards while still turning away penetration testing and other techniques that lean toward attacker use.

Its biology safeguards match those on Sonnet 5, Sonnet 5.5, and Opus 5: research-level questions pass, while requests judged likely to cause harm are held back.

Claude Haiku 5.5 is available today with experimental support to Kiro Pro, Pro+, Pro Max, and Power customers in us-east-1 (N. Virginia) and eu-central-1 (Frankfurt). It ships with a 0.06x credit multiplier. Requests that exceed 100K tokens are billed at a 0.3x multiplier, consistent with Anthropic’s and Amazon Bedrock’s long context pricing for Haiku 5.5.