Key takeaways:
- Anthropic launches Claude Haiku 5.5, its third Claude 5.5 model.
- Haiku 5.5 costs 75% less than its predecessor, Haiku 4.5.
- The model adds cybersecurity safeguards while targeting high-volume AI tasks.
Anthropic launched Claude Haiku 5.5 on Wednesday, its third Claude 5.5 model in a month, targeting high-volume AI tasks with lower costs as the company prepares for a planned initial public offering.
Haiku 5.5 cuts AI costs
Haiku 5.5 is designed for tasks including classification, summarization, data extraction and other repetitive workloads. Anthropic also said the model is suited for live customer support, voice agents and assistants built into applications.
The company said Claude Haiku 5.5 costs about 75% less to run than its predecessor, Haiku 4.5. For prompts of up to 100,000 tokens, Anthropic prices the model at 10 cents per million input tokens and 50 cents per million output tokens.
For prompts above 100,000 tokens, the price rises to 50 cents per million input tokens and $2.50 per million output tokens. Anthropic said about 90% of requests to its previous Haiku model were within the lower prompt range.
Anthropic also cut the price of cached inputs for Claude Sonnet 5.5 by 50%, saying the change reduces the cost of most agent-based work with the model by about 20%.
Anthropic adds safety controls
Haiku 5.5 includes tighter safeguards for cybersecurity requests than Haiku 4.5, while allowing a wider range of defensive security work than Anthropic’s safeguards for Sonnet 5.5.
The company said the safeguards block penetration testing and other techniques that are more likely to be used by attackers. Anthropic has separately expanded its Cyber Verification Program to give qualifying security teams access to more advanced models under different levels of safeguards.
Anthropic said Haiku 5.5 also shows improvements in its alignment evaluations compared with Haiku 4.5, including fewer instances of behavior the company considers misaligned and less willingness to cooperate with misuse.
Alex Wang, an applied AI executive at Anthropic, said the model is intended for high-volume work where speed and cost matter.
At HubSpot, Distinguished Software Engineer Ze’ev Klapow said the company’s testing found Haiku 5.5 produced its strongest results yet on a CRM evaluation suite, averaging a 92.8% score across three runs. Anthropic published the comments as part of its launch announcement.
Third Claude 5.5 model arrives
Haiku 5.5 follows Claude Opus 5.5, released Sept. 22, and Claude Sonnet 5.5, released Sept. 28. Anthropic has therefore added three models to its Claude 5.5 family in less than a month.
The releases come as Anthropic builds its business ahead of a planned IPO. The company’s rapid product rollout also comes amid intense competition among AI developers to improve model performance while reducing the cost of running AI systems.
Haiku 5.5 is available through Anthropic’s Claude platform as well as Amazon Web Services, Google Cloud and Microsoft Azure. It is also available through Claude Code.
