Claude Haiku 5.5 Launch: Lower AI Costs and What They Mean
AI UPDATE · OCTOBER 12, 2026
The short version: Anthropic introduced Claude Haiku 5.5 on October 7, 2026. The new model targets people building fast, high-volume AI features where the cost of every request matters, including summarization, classification and support workflows.
What is Claude Haiku 5.5?
Haiku is Anthropic’s lighter-weight model family. Haiku 5.5 is positioned for shorter, repeated tasks, while Anthropic continues to position larger Sonnet and Opus models for harder workflows. Examples Anthropic highlights include document summaries, sorting incoming requests, pulling useful context into a larger system and supporting coding agents.
The company also introduced an adjustable effort setting for Haiku 5.5, letting developers choose how much computation to spend on different requests. That flexibility is important when an application processes thousands of similar tasks instead of one complex problem.
Pricing: what changed?
According to Anthropic’s launch pricing, standard API input for prompts up to 100,000 tokens costs $0.10 per million tokens and output costs $0.50 per million tokens. Longer prompts have different rates, so developers need to check the length of their requests and the latest billing documentation rather than assuming a single flat price.
Anthropic estimates typical Haiku 5.5 workloads cost about 75% less than Haiku 4.5 on average. That figure comes from Anthropic’s own comparisons and depends on prompt length, token use and workload. It should not be treated as a guaranteed saving for every application.
Where is it available?
Anthropic lists availability through Claude’s own services and developer platform, as well as supported cloud providers including Amazon Web Services, Google Cloud and Microsoft Azure. Developers can identify the model on the Claude Platform using claude-haiku-5-5.
Why it matters
For businesses that use AI in repeated small jobs, efficiency can matter as much as maximum capability. A model that is fast and inexpensive enough for large volumes might make background features—like sorting tickets or summarizing notes—more practical.
But a cheaper model is not automatically the right model. Longer, complex coding tasks and high-stakes work may still need a stronger model, human verification, or both. Anthropic’s own comparisons say Sonnet and Opus remain better suited to some demanding agentic tasks. Before switching, teams should compare error rates, latency, privacy requirements and total cost on their own workload.
Sources and context
This article reports Anthropic’s published product information. Pricing and availability can change; USABite has not independently benchmarked Haiku 5.5.
