Highlights:

  • The company claims Haiku is three times faster than competitors, which is ideal for high-speed, low-latency use cases.
  • Anthropic states that Haiku can process up to 21,000 tokens, approximately equivalent to 30 pages of text per second for prompts under 32,000 tokens.

An artificial intelligence startup comparable to OpenAI’s GPT-4 and specializing in developing reliable AI models, Anthropic PBC launched Claude 3 Haiku recently. This latest addition joins the Claude 3 family of models prioritizing speed and cost-effectiveness.

In March, Anthropic unveiled the Claude 3 series of large language models comprising three variants. The company asserts that its leading model, Claude 3 Opus, boasts substantial processing capabilities, competing closely with industry leaders like OpenAI and Google LLC. Meanwhile, its counterpart, Sonnet, prioritizes speed for cost efficiency.

According to the company, Haiku outpaces its counterparts by threefold in processing most workloads, rendering it particularly suitable for scenarios demanding rapidity and minimal latency. This makes it well-suited for applications like customer service, fieldwork, question-and-answer systems, and other instances where swift responses are crucial.

“Speed is essential for our enterprise users who need to quickly analyze large datasets and generate timely output for tasks like customer support. It also generates swift output, enabling responsive, engaging chat experiences and the execution of many small tasks in tandem,” the company stated in the announcement.

Anthropic claims that Haiku can process up to 21,000 tokens, equivalent to approximately 30 pages of text, per second for prompts under 32,000 tokens.

Similar to the other models within the Claude 3 series, Haiku can address fundamental inquiries and requests. With a maximum prompt size of 200,000 tokens, equivalent to about 150,000 words or over 500 pages of content, Haiku offers expanded capabilities. The company stated that all three models excel in content creation, code generation, and analysis and demonstrate improved fluency in non-English languages such as Spanish, Japanese, and French.

The company also emphasized affordability by implementing a 1:5 pricing model based on the input-to-output token ratio for enterprise workloads, where longer prompts are typical. Businesses frequently depend on Large Language Models (LLMs) to process and analyze extensive documents, which can result in elevated expenses. Anthropic indicated that for just USD 1, the model could analyze 400 Supreme Court cases or 2,500 images.

The company said, “Businesses can rely on Haiku to quickly analyze large volumes of documents, such as quarterly filings, contracts, or legal cases, for half the cost of other models in its performance tier.”

As revealed in the announcement, Anthropic announced that Haiku will be joining Sonnet on Amazon Web Service Inc.’s public cloud via Amazon Bedrock, a managed service offering access to AI foundation models from AWS and various other companies. Additionally, the company stated that the model will soon be available on Google Cloud Vertex AI, a platform by Google LLC designed for training and deploying generative AI models.

Both customers and developers can choose to utilize Haiku either through the company’s application programming interface or by subscribing to Claude Pro via claude.ai.