Anthropic Launches Claude Haiku 5.5 Small Model

Anthropic introduced Claude Haiku 5.5 on October 7, 2026, positioning it as the cheapest and fastest small model it has shipped, intended for high volume, cost sensitive work rather than complex reasoning. The company said the model handles repetitive tasks such as summaries, compactions, database queries and classification, and that it pairs with Sonnet 5.5 and Opus 5.5 as a subagent on coding tasks. It is also Anthropic’s first Haiku class model with an adjustable effort setting, letting a developer trade intelligence for cost on the same model rather than switching between two.

On pricing, Anthropic published a table showing Haiku 5.5 costing $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, rising to $0.50 and $2.50 above that threshold. Those figures compare against $1.00 and $5.00 for Haiku 4.5. Anthropic noted that roughly 90% of requests to the previous Haiku model fell under the 100,000 token threshold, which is why it frames the headline reduction as applying to most real traffic rather than a narrow slice of it.

Alongside the release, Anthropic cut Sonnet 5.5’s cache read price in half, from $0.20 to $0.10 per million tokens, saying that because cache reads make up a large share of token consumption on agentic tasks, the change lowers the cost of running Sonnet 5.5 on most agentic work by around 20%. The company also said it would begin rolling out monthly API credits this week to Claude Max and Team subscribers, letting them experiment with building tools and agents on the Claude Platform at no added charge up to the credit amount.

Anthropic published benchmark results comparing Haiku 5.5 against Haiku 4.5, GPT-6 Luna and Sonnet 5.5 across tasks including computer use, knowledge work and agentic coding, and said the model carries more restrictive cybersecurity safeguards than Haiku 4.5 though less restrictive than those applied to its newest large models. The model is available now through the Claude Platform and through Amazon Web Services, Google Cloud and Microsoft Azure.

The numbers

• Haiku 5.5 costs around 75% less to run than Haiku 4.5 on average, Anthropic said.
• Sonnet 5.5 cache reads were cut 50%, from $0.20 to $0.10 per million tokens.
• That cache read cut lowers Sonnet 5.5’s cost on most agentic work by around 20%.
• About 90% of requests sent to the previous Haiku model used prompts under 100,000 tokens, the range where the steepest price cut applies.