The new Anthropic Ctations feature aims to reduce AI errors

In an announcement that may have been placed to undermine the interest of OpenAI DriverAnthropic Thursday was unveiled a new feature of its editing API called Citationswhich allows for “ground” feedback from its own Claude the AI ​​family in basic documents such as emails.

Anthropic says Citations allows its AI models to provide detailed references to “specific sentences and paragraphs” from the text it uses to generate answers. As of Thursday afternoon, Articles are available in Anthropic’s API and Google’s Vertex AI platform.

As Anthropic explains in a post blog with Citations, devs can add source files to include only citation types that claim to have referenced the files. Text is especially useful for document condensing, Q&A, and customer service programs, Anthropic says, where the feature can move samples to insert text.

Text is not available for all Anthropic versions – Claude 3.5 Sonnet and Claude 3.5 Haiku only. Also this feature is not free. Anthropic states that Citations may be reimbursed depending on the length and number of articles cited.

Based on Anthropic’s API pricing, which Citations uses, a basic doc of about 100 pages would cost $0.30 with a Claude 3.5 Sonnet, or $0.08 with a Claude 3.5 Haiku. This can be beneficial for devs who want to minimize it’s a nightmare and other errors caused by AI.