Anthropic released Claude Haiku 5.5 on October 7, 2026, adding a new small and fast model to the Claude family. The company says the model is designed for high-volume, cost-sensitive workloads, including summarization, context compaction, database queries, classification, customer support and browser use where response speed matters. Reuters reported that Haiku 5.5 is the third model released in the Claude 5.5 family.
Pricing uses two context tiers
According to Anthropic’s published pricing table, prompts of up to 100,000 tokens cost $0.10 per million input tokens and $0.50 per million output tokens. In the same tier, cache reads cost $0.01 per million tokens and cache writes cost $0.125. For prompts above 100,000 tokens, input rises to $0.50, output to $2.50, cache reads to $0.05 and cache writes to $0.625 per million tokens.
Anthropic says the average cost per workload is about 75% lower than Haiku 4.5 after accounting for differences in token usage. In a footnote, the company says list pricing is 90% lower for requests up to 100,000 tokens and 50% lower for longer requests. Those figures are the vendor’s own pricing comparison, not an independent cost audit, and actual spending will depend on prompt length, generated output, caching and retry rates.
Adjustable effort and a speed-focused role
Haiku 5.5 is the first Haiku-class model to include an adjustable effort setting. Developers can use that control to balance cost against the amount of computation assigned to a task. Anthropic positions the model for narrow, repetitive jobs and as a subagent working alongside larger models. It also says Sonnet 5.5 and Opus 5.5 remain better choices for complex, long-running agentic coding tasks.
The company published benchmark charts and early customer reports that indicate gains in speed and task performance. These results should be read with appropriate context: they were selected and presented by Anthropic and do not amount to an independent benchmark review. Performance in production will vary with the application, the size and quality of the input, tool use and the evaluation criteria chosen by each customer.
Available across major cloud platforms
Claude Haiku 5.5 is available on the Claude Platform under the model identifier claude-haiku-5-5. Anthropic also says the model is available through Amazon Web Services, Google Cloud and Microsoft Azure. That distribution gives enterprise customers the option to evaluate the model within existing cloud contracts, security controls and procurement processes.
Alongside the launch, Anthropic cut the price of Sonnet 5.5 cache reads from $0.20 to $0.10 per million tokens. The company estimates that change will reduce the cost of Sonnet 5.5 by about 20% for most agentic workloads. It also announced monthly API credits for subscribers: $100 for Max 5x users, $200 for Max 20x users and up to $500 pooled across a Team account.
The release illustrates how smaller AI models are being positioned not only as cheaper chat systems but also as specialized components that handle frequent, bounded tasks inside larger applications. The headline price advantage, however, is not a guarantee of total application savings. Developers will still need to measure latency, output quality, context size and the number of retries required for their own workloads before deciding whether the model lowers end-to-end cost.
