Anthropic introduces Claude Haiku 5.5 model

Nearly a year after its previous version, Anthropic has introduced the Claude Haiku 5.5 model. The new model is designed for fast performance and relatively low-cost usage.

According to the company, Claude Haiku 5.5 was developed for short and repetitive tasks, data classification, and programming-related processes. The model aims to improve the ratio between speed and price.

In results presented by Anthropic, Claude Haiku 5.5 scored higher than GPT-6 Luna in certain tests. Specifically, in the OSWorld 2.1 benchmark, the new model showed a result of 72.4 percent, while GPT-6 Luna recorded 48.9 percent.

Claude Haiku 5.5 also achieved a high result in the GDPval-AA test. In this benchmark, Haiku 5.5 scored 1,620 points, whereas GPT-6 Luna obtained 1,437 points.

The pricing for Claude Haiku 5.5 is set at $0.10 per 1 million input tokens and $0.50 per output token. Anthropic stated that compared to Claude Haiku 4.5, costs have been reduced by up to 75 percent.

Thus, through the new model, Anthropic aims to strengthen competition in the segment of tasks requiring fast performance and low costs.