Claude Haiku 5.5: Features, Pricing and How to Use It

What is Claude Haiku 5.5?

Claude Haiku 5.5 is a fast, low-cost small artificial intelligence model developed by Anthropic and released on 7 October 2026. This model is designed specifically to handle quick, repetitive, and cost-sensitive tasks at scale.

Compared to its predecessor, Claude Haiku 4.5, this updated version is significantly more capable and runs at a much lower cost. According to Anthropic, running the new model is about 75 percent cheaper on average.

What can Claude Haiku 5.5 do?

Claude Haiku 5.5 excels at high-volume, speed-sensitive applications such as live customer support, database queries, summaries, and classification requests. It is also designed to act as an efficient subagent alongside larger models like Claude Sonnet 5.5 and Claude Opus 5.5 to perform specific tasks on coding projects.

This is the first Haiku-class model to feature an adjustable effort setting, letting users balance cost and intelligence. Users can choose between low, medium, high, xhigh, and max reasoning effort levels, though reasoning cannot be disabled entirely and defaults to medium.

In early evaluations shared by Anthropic, corporate users reported notable performance gains. HubSpot recorded a 92.8 percent average score on CRM deal reporting, while Box noted that the model completed analytical tasks with roughly half the latency of Claude Haiku 4.5.

How much does Claude Haiku 5.5 cost?

Claude Haiku 5.5 matches the pricing of competing small models like OpenAI's GPT-6 Luna, costing $0.10 per million input tokens and $0.50 per million output tokens. However, this competitive rate only applies to prompts containing up to 100,000 tokens.

For tasks exceeding 100,000 tokens, the cost of Claude Haiku 5.5 increases fivefold to $0.50 per million input tokens and $2.50 per million output tokens. Additionally, the model utilizes a less generous tokenizer that requires roughly 1.25 times more tokens for the same long prompt compared to Claude Haiku 4.5.

How to use Claude Haiku 5.5

Claude Haiku 5.5 is available across multiple cloud platforms and directly through the Claude API. Developers can access the model using Amazon Web Services, Google Cloud, Microsoft Azure, or the Claude Platform.

1. Install or update the Anthropic command-line tool plugin: llm install -U llm-anthropic
2. Refresh your active models to fetch the new release: llm anthropic refresh
3. Run a test query with your preferred effort setting: llm -m claude-haiku-5.5 "Your prompt here" -o thinking_effort low
4. To use monthly API credits as a Claude Max or Team subscriber, navigate to Settings, select Billing, and assign the credit to your chosen API organization.

What are the limits of Claude Haiku 5.5?

Claude Haiku 5.5 is subject to strict safety and alignment guardrails set by Anthropic. While the cybersecurity safeguards are more permissive than those on Claude Sonnet 5.5 to allow defensive operations, the model still blocks penetration testing and other offensive techniques.

Biology safeguards restrict requests that could cause physical harm while still permitting basic research biology questions. Furthermore, developers working with large volumes should monitor context limits closely, as prompts longer than 100,000 tokens trigger the 5x pricing multiplier.

Frequently asked questions

Sources: Anthropic · Simon Willison