Claude Haiku 5.5 lands on Amazon Bedrock
A 75% price cut and new effort controls, built to run as a subagent under Opus 5.5.
The facts
- Claude Haiku 5.5 is now available on Amazon Bedrock and Claude Platform on AWS
- Anthropic says it costs around 75 percent less than Claude Haiku 4.5 for most tasks
- It is the first Haiku model with effort controls, letting developers tune cost against intelligence per task
- Available via US, EU, AU, JP and Global CRIS inference profiles on bedrock-runtime, plus GovCloud (US) on bedrock-runtime and bedrock-mantle endpoints
Amazon Bedrock has added Claude Haiku 5.5. Anthropic describes it as the fastest and most efficient model in its Claude 5.5 family. AWS says it is built for subagents and high-volume, cost-sensitive work. It costs around 75 percent less than Claude Haiku 4.5 for most tasks, according to Anthropic.
The model is Anthropic's most capable Haiku yet across coding, tool use, computer use and agentic tasks, AWS reports. It is also the first Haiku model with effort controls. These let developers dial cost against intelligence per task, rather than setting one level for an entire workload. AWS points to use cases including routing requests, reviewing code, classifying long documents, scanning knowledge bases, and running as a computer-use subagent for repetitive browser and desktop work. It also handles agentic coding, multi-step tool use and high-resolution images.
AWS is positioning Haiku 5.5 as the fast half of a two-model pairing with the recently announced Claude Opus 5.5. Opus plans and makes judgment calls on hard reasoning, such as release debugging or security review of large pull requests. Haiku executes high-volume, well-defined work instead: classifying, summarising, rewriting documents, and applying small changes across many files. Haiku is cheap and fast, AWS says, so teams can run many of its subagents in parallel under a single Opus coordinator.
On Bedrock, Haiku 5.5 runs under IAM access controls, CloudTrail audit logging, CloudWatch monitoring and Bedrock Guardrails, with usage billed through AWS. It's accessible today through US, EU, AU, JP and Global CRIS inference profiles on bedrock-runtime. It is also available through both bedrock-runtime and bedrock-mantle in AWS GovCloud (US). A further route is Claude Platform on AWS in North America, which gives direct access to Anthropic's native platform experience, unified with AWS billing and authentication. AWS did not publish specific per-token pricing in this post. It points instead to the Amazon Bedrock pricing page.
Why it matters
The price cut and per-task effort controls make Haiku the default choice for teams fanning out many parallel subagents under a single Opus orchestrator. That cost structure favours agentic architectures over single large-model calls. Anyone running Haiku 4.5 at scale on Bedrock should re-run cost projections before the next budget cycle.
Primary source
AWS Machine Learning — aws.amazon.comWritten from the primary source above by The Neural Net’s automated desk (claude-sonnet-5, theneuralnet/pipeline@1) without human review. How this works. Corrections: check the source first — if we got it wrong, the source link is the record.