Anthropic Launches Cheap and Fast Claude Haiku 5.5 Ahead of Planned IPO

Anthropic introduced Claude Haiku 5.5 on Wednesday, October 8, marking its third model release in a month. The company describes it as its cheapest, fastest, and most capable small model built to handle high-volume and cost-sensitive tasks.

Performance And Pricing

Haiku 5.5 manages quick and repetitive workloads such as summaries, database queries, and classification requests. Anthropic states that the model costs around 75 percent less to run than Haiku 4.5. It is priced at $0.10 per million input tokens and $0.50 per million output tokens. This is significantly lower than Haiku 4.5, which costs $1 and $5, and Sonnet 5.5, which costs $2 and $10 for the same input and output amounts.

Benchmarks And Features

The model secured a score of 1620 on the GDPval-AA v2.1 knowledge work test, beating OpenAI's GPT-6 Luna, which scored 1437. On the OSWorld 2.1 computer use test, Haiku 5.5 achieved 74 percent compared to GPT-6 Luna's 48.9 percent. It also scored 45.9 percent in Humanity's Last Exam. The model is the first in the Haiku class to feature an adjustable effort setting, letting users choose between cost and intelligence.

Context

AI labs frequently release smaller, faster models to help businesses cut down on computing costs for daily tasks. This particular release comes as Anthropic prepares for a potential stock market launch through an initial public offering.