According to Anthropic, Claude 3’s Haiku family member has the fastest genetics and is available at an affordable price.

At the beginning of March, Anthropic released the new Claude 3 family, starring Sonnet, Opus and Haiku, next in the series after Claude 2.1. Like the other models in the Claude 3 family, Haiku can respond to basic questions or requests. What sets Haiku apart, according to Anthropic, is its speed and affordability. Users with a Claude 3 Pro subscription can already test the model.

The fastest in the family

The Claude 3 family consists of Sonnet, Opus and Haiku. The latter is the fastest of the three, according to Anthropic, a claim the company tries to prove with some benchmarks on its blog. According to the AI ​​company, Haiku can process up to 21,000 tokens per second, which is equivalent to about 30 pages of text.

According to Anthropic, Haiku is said to be three times faster than its competitors when it comes to handling most workloads. This makes the model suitable for use cases that require speed and low latency, such as customer service, Q&A, or field service.

Although the company has overtaken OpenAI, the ChatGPT developer seems to be catching up and the first signals of a GPT 4.5 turbo are visible.

Affordable price

In addition to speed, Anthropic also placed great emphasis on price when developing the Haiku model. To achieve this, the company has introduced a 1:5 input-to-output token ratio pricing model for enterprise workloads where longer prompts are common. Most companies that use LLMs process and analyze large documents, which can result in high costs.

Haiku would join Sonnet on Amazon Web Service Inc.’s public cloud. via Amazon Bedrock. In addition, the model will soon also be available for Google Cloud Vertex AI. Customers and developers can now use the model with a Claude Pro subscription via claude.ai.